Documente Academic
Documente Profesional
Documente Cultură
IN METALLURGY
G. H. GEIGER
University of Illinois, Chicago Circle
D.R. POIRIER
University of Bridgeport
...
'Y'Y
i\DDISON-WESLEY PUBLISHING COMPANY
Reading, Massachusetts Menlo Park, California
London Amsterdam Don Mills, Ontario Sydney
FOREWORD
Despite the fact that metallurgists and metallurgical engineers have been dealing
with processing problems and metallurgical reactions for many years, there has
not been any text available for metallurgical educators who would like to present
their students with a unified fundamental approach to fluid dynamics, heat
transfer, and mass transfer. Also, the remarkable advances made in the field of
metallurgy during the past decade have mostly been limited to the scientific areas,
while the engineering aspects of metallurgical engineering have largely been
neglected. This book provides an up-to-date approach to transport phenomena
and then goes further by discussing numerous examples, applications, and
phenomena which are of particular importance to metallurgists.
Quite a few books on transport phenomena (momentum, heat, and mass
transport) have appeared over the past decade; some are intended for chemical
engineers and others for mechanical engineers. When we started to organize and
teach a course to juniors in metallurgical engineering, we found that these were
only partly satisfactory, mainly for pedagogical reasons. Students of metallurgy,
when exposed to the subject matter of transport phenomena without benefit of a
substantial effort to i~troduce them to the direct link between transport phenomena
and metallurgy, could hardly gain an appreciation of the fact that transport
phenomena are just as basic to the field of metallurgy as is physical metallurgy.
We also felt the need for a text that would utilize and match the mathematical
abilities of upper-class students o( metallurgical engineering and metallurgy.
Most upper-class studen1s in metallurgical engineering have studied integral and
differential calculus, but often make no use of it again, except in the most trivial
applications. Since one can proceed quite a way into the field of transport
phenom<:<na without resorting to higher-level mathematics, and because we think
that it is important for the student, and certainly for the practicing engineer, to
visualize the physical situations, we have attempted to lead the reader through the
development. and solution of the necessary differential equations in a manner as
clear as possible, by applying the familiar principles of conservation to numerous
situations and by purposefully including many intermediate mathematical steps.
We hope that students with good mathematical abilities will not feel that we are
being condescending, but will realize that we are trying to make the material
vii
understandable to a group as wide as possible. The reader who is not "up" on carburizing and decarburizing of single- and two-phase steels, homogenization of
his mathematics may still learn from the book, but must pay careful attention to cast and banded structures, and the diffusion of gases into or out of basic simple
the characteristics of each situation and the boundary conditions for which a shapes. Interphase and interfluid reactions are also discussed and the concept of
particular solution is valid in order to avoid applying a wrong solution, which the "controlling step" is introduced and illustrated by examples such as a discus-
unfortunately happens all too often. sion of the reduction of iron oxide, metal vapor loss from melts under vacuum,
The book is organized in the seemingly traditional manner characteristic of and many others.
texts in transport phenomena; Section I deals with the properties and mechanics The book may be used as a supplement to several courses besides a primary
of fluid motion; Section II with thermal properties and heat transfer; and course in transport processes. For example, by starting with Chapters 6, 7, and 9,
Section III with mass transfer. However, we have departed from the tradition in Chapter IO may be taught as a part of a course in solidification or foundry
several significant ways. engineering. One of the authors has regularly included material from Chapters
The first chapter of each section is devoted to transport properties of sub- 3, 5, 12, 15, and 16 in courses in process or extractive metallurgy. Chapters 13, 14,
stances which are of interest to metallurgical engineers. In all eases, the transport 15, and 16 should make a good basis for a course in heterogeneous kinetics. The
properties of materials are discussed from a structural point of view; in this material in Chapters 6, 7, 8, 9, 11, and 14 has been satisfactorily used, following a
regard, some of the material is easily recognizable by physical metallurgists, course in fluid mechanics, as the basis for part of a course in heat treating.
particularly in the discussions of diffusion coefficients in solids. On the other We would like to emphasize that the book is not intended to be a comprehen-
hand, the viscosity of liquid metals and molten slags and salts, which is dealt with sive review of all the applications and current research in the field. Its aim is
in the book, is hardly ever touched on in most curricula throughout this country. rather to present the basic equations and show how they can be applied to a
Viscosity is a structure-sensitive property and, as such, should be of interest to all variety of topics, thereby preparing the reader to make direct use of the information
metallurgists. To mention a few aspects of the way thermal conductivity is treated, or go on to develop their applications, and also enabling him to read the current
there are the usual basic discussions of gases and liquids and also a discussion of literature in the field. Each chapter is followed by Problems; many of them are
pure single-phase metals leading to the Lorenz number. But, beyond that, some numerical and we have tried very much to gather problems that not only have
attention is given to two-phase solids, and bulk, porous materials, such as molding learning value but also a factor of interest catering to students of metallurgy. We
sand for castings, are discussed in great detail. hope that the book will be found useful and will help metallurgists improve their
While the first section of the book dealing with fluid dynamics points out many analyses and also provide aid to those who already are actively participating in
concerns of the process metallurgist, such as packed beds of solids, fluidized beds, trying to find better ways how to engineer their processes.
flow through pipes, meters, and other phenomena, and then gives an introduction
to the engineering use of fans, blowers, high-velocity jets and vacuum systems, it Chicago, Illinois G.H.G.
should also be of interest to the physical metallurgist, mentioning such topics as Bridgeport, Connecticut D.R.P.
elutriation characterizing of particles, and powder rolling, both of importance in April 1972
powder metallurgy. In addition, one need only scan through the recent metal-
lurgical literature to see that convection is a very important part of solidification
phenomena, traditionally of prime interest to physical metallurgists.
The subject matter of heat transfer included herein covers the fundamentals of
heat transfer by conduction and radiation, and heat transfer with convection.
Going further, so as to stimulate metallurgists and metallurgical engineers,
numerous applications to heating and cooling processes are discussed. Quenching,
heating of billets, heating and cooling of strip by radiation and sprays are some
examples. Chapter IO is devoted to solidification heat transfer and Chapter 12 to
heat transfer in packed beds, imparting definite metallurgical flavor to the text.
The section on mass transfer is generous in the portion devoted to diffusion in
solids (metals and ionic solids as well). Mechanisms bf diffusion in solids are
discussed, and diffusion of metal vapors in gases at high temperatures is treated.
Drawing heavily on the mathematical similarity between conduction heat transfer
and diffusion in solids, numerous diffusional problems are discussed including
ACKNOWLEDGMENTS CONTENTS
.' 'F)-le authors express their sincere thanks to Professor Richard' W. Heine of the PART 1 FLUID DYNAMICS
University of Wisconsin for his encouragement and support of our effort, initiated
while we were in Madison. We )VOuld also like to especially thank our former fl.Chapter 1 Properties of Fluids
mentors, Professor J. Bruce Wagner of Northwestern University and Professor 1.1 Types of fluid flow 3
Merton C. Flemings of the Massachusetts Institute of Technology, for their 1.2 Newtonian fluids 5
encouragement and guidance when we pursued our graduate studies; their 1.3 Viscosity of gases 7
influence is particularly recognizable in Chapters 10 and 14. Many of our students 1.4 Viscosity of liquids 13
who had to put up with the inconvenience of studying our notes and rough manu- 1.5 Non-Newtonian fluids 31
scrip~ deserve our thanks and belated sympathies. They had the thankless job of
helping us to sift and sort out so that finally the manuscript could be completed. Chapter 2 Laminar Flow and Momentum Balances
We also acknowledge our indebtedness to the pioneering contributors to the 2.1 Momentum balance 36
field of transport phenomena; noteworthy among these are Professors R. Byron 2.2 Flow of a falling film 37
Bird, Warren E. Stewart, and Edwin N. Lightfoot of the University of Wisconsin. 2.3 Flow between parallel plates 41
They have given permission to use a number of figures, tables, and examples from 2.4 Flow through a circular tube 43
their classic text, Transport Phenomena. And it was they who, as much as anybody, f- 2.5 General momentum equations. . . . . . . . . . 47
have pointed out that "the subject of transport phenomena should rank along with 1" 2.6 The conservation of momentum equation in curvilinear coordinates 53
thermodynamics, mechanics, and electromagnetism as one of the key engineering 'i1 2.7 Application ofNavier-Stoke's equation . 59
sciences." We are also grateful to Professor John F. Elliott for critically reading
the manuscript and for his gracious Foreword to this book, to Professor C. Wagner i Chapter 3 Turbulent Flow and Experimental Results
for permission to use material from his notes presented for many years at M.l.T., 3.1 Friction facfors for flow in tubes 75
and to Professor Morris Cohen of M.I.T. for his early endorsement of the 3.2 Flow in noncircular conduits . 82
manuscript. 3.3 Flow past submerged bodies . 83
Finally we owe a debt of gratitude to our wives who, after earning their Ph.T. 3.4 Flow through packed beds of solids 91
degrees (putting husband through), had to listen to all our gripes directed to this 3.5 Fluidized beds . 102
****** book; to our children, who were constantly bullied away from the kitchen Chapter 4 Energy Balance Applications in Fluid Flow
table and "shh-shh-ed" so that we could discuss the manuscript; and to our parents
who have promised to buy a copy of the book even if no one else does. 'f 4.1 Conservation of energy. 111
"*4.2 Friction losses in straight conduits 114
4.3 Enlargement and contraction . . 117
4.4 Flow through valves and fittings . 118
4.5 Flow through smooth bends and coils 121
4.6 Flow measurement . ' 123
4. 7 Flow from ladles 134
xi
xn Conti
*
Chapter 8 Correlations for.Heat-Transfer with Convection
Chapter 13 Fick's Law and Diffusivity of Materials
8.1 Heat-transfer coefficients for forced convection in tubes 242
249 13. l Definition of fluxes-Fick's first law 431
8.2 Heat-transfer coefficients for forced convection past submerged objects
252 13.2 Diffusion in solids . 433
8.3 Heat transfer coefficients for natural convection.
256 13.3 Diffusion in solid nonmetals 449
8.4 Quenching heat-transfer coefficients
13.4 Diffusion in liquids . 455
8.5 Boiling heat transfer 263
13.5 Diffusion in gases . 463
~ Chapter 9 Conduction of Heat in Solids
13.6 Diffusion through porous media 467
xv
--------------------
Figure raoieSource Page
Figure-Table Source Page
Tables R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phe- 54-58
Fig. 5.18 Adapted from A. R. Anderson and F. R. Johns, Jet Propulsion 25, 163
2.l-2.1 nomena, Wiley, New York, 1960, pages 83-91. Reprinted by per-
13 (1955).
mission of John Wiley and Sons, Inc.
Fig. 5.19 G. C. Smith, J. Metals, 846 (July, 1966). By permission of The 164
Fig. 2.7 L. Howarth, Proc. Roy. Soc., London Al64, 547 (1938). 64
Metallurgical Society of AIME.
Fig. 3.2 Adapted from L. F. Moody, Trans. ASME 66, 671 (1944), and 80
Fig. 5.20 Adapted from A. R. Anderson and F. R. Johns, Jet Propulsion 25, 164
Mech. Eng. 69, 1005 (1947).
13 (1955).
Fig. 3.3 W. M. Rohsenow and H. Y. Choi, Heat, Mass, and Momentum 83
Fig. 5.21 Adapted from A. R. Anderson and F. R. Johns, Jet Propulsion 25, 165
Transfer, Prentice-Hall, Englewood Cliffs, New Jersey, 1961,
13 (1955).
page 63.
Table 5.2 J. M. Lafferty, Techniques ofHigh Vacuum, General Electric Report 171
Fig. 3.5 J. K. Yennard, Elementary Fluid Mechanics, Wiley, New York, 84
No. 64-RL-3791G, 1964.
1954, page 384. Reprinted by permission of John Wiley and Sons,
Fig. 5.30 From an article by F. Berkeley, excerpted by special permission 179
Inc.
from Chem. Eng., April, 1957, page 255, by McGraw-Hill, Inc.,
Fig. 3.8 Adapted from F. Eisner, Proc. 3rd Intern. Congr. Appl. Mech., 1930, 88
New York, N.Y. 10036.
page 32.
Fig. 6.3 A. Schack, Industrial Heat Transfer, Wiley, New York, 1965. 189
Fig. 3.9 Adapted from F. Eisner, Proc. 3rd Intern. Congr. Appl. Mech., 1930, 90
Fig. 6.4 W. J. deHaas and T. Biermasz, Physica 2, 673 (1935); 4, 752 (1937); 190
page 32.
5, 47, 320, and 619 (1938).
Fig. 3.10 Based on S. Ergun, Chem. Eng. Prog. 48, 93 (1952). 95
Fig. 6.7 A. Schack, Industrial Heat Transfer, Wiley, New York, 1965. 194
Table 3.2 Tabular values of shape factors calculated from data presented by 96
Fig. 6.10 D. Nobili and M.A. DeBacci, J. Nuclear Materials 18, 187 (1966). 199
P. C. Carman, Trans. Inst. of Chem. Eng. 15, 155 (1937).
By permission of North-Holland Publishing Co.
Fig. 3.11 Derived by H. E. Rose and A. M. Rizk, Proc. Inst. Mech. Engrs., 99
Fig. 6.11 P. Grootenhuis, R. W. Powell, and R. P. Tye, Proc. Phys. Soc. 65, 200
London 60, 493 (1950).
502 (1952). By permission of The Institute of Physics and The
Fig. 3.12 C. C. Furnas, Ind. and Eng. Chem. 23, 1052 (1931). Reprinted by 102
Physical Society.
permission of the American Chemical Society.
Fig. 6.12 R. G. Deissler and C. S. Eian, Nat. Advisory Comm. Aeronaut., 202
Fig. 3.13 C. C. Furnas, Ind. and Eng. Chem. 23, 1052 (1931). Reprinted by 102
RM E52C05 (1952).
permission of the American Chemical Society.
Fig. 6.13 Experimental values from L. F. Lucks, C. L. Linebrink, and K. L. 205
Table 13.3 H. White and S. Walton, J. Am. Ceramic Soc. 20, 155 (1937). 103
Johnson, Trans. American Foundrymen's Society 55, 62 (1947).
Fig. .3.14 C. Y. Wen and Y. H. Yu, Fluid Particle Technology, Chem. Eng. 104
Table 7.1 W. M. Rohsenow and H. Y. Choi, Heat, Mass, and Momentum 212
Prog. Symposium Series, No. 62, AIChE, New York, 1966.
Transfer, Prentice-Hall, Englewood Cliffs, New Jersey, 1961,
Fig. 3.16 C. Y. Wen and Y. H. Yu, Fluid Particle Technology, Chem. Eng. 106
page 141.
Prog. Symposium Series, No. 62, AIChE, New York, 1966.
Fig. 7.4 E. Z. Pohlhausen, Zf angew. Math. u. Mech. l, 115 (1921). 214
Fig. 4.3 W. M. Kays and A. L. London, Trans. ASME 14, 1179 (1952). 118
Fig. 7.7 Calculated by S. Ostrach, Nat. Advisory Comm. Aeronaut. Tech. 224
Fig. 4.4 W. M. Kays and A. L. London, Trans. ASME14, 1179 (1952). 119
Note 2635, Feb., 1952, as presented in W. M. Rohsenow and H. Y.
Fig. 4.6 Steam-Its Generation and Use, Babcock and Wilcox, New York, 120
Choi, Heat, Mass, and Momentum Transfer, Prentice-Hall, Engle-
1963, pages 8-17.
wood Cliffs, New Jersey, 1961, pages 155-159.
Table 4.2 W. M. Rohsenow and H. Y. Choi, Heat, Mass, and Momentum 120
Tables R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phe- 233
Transfer, Prentice-Hall, Englewood Cliffs, New Jersey, 1961, 120
7.3-7.5 nomena, Wiley, New York, 1960, pages 317-319. Reprinted by
page 64.
permission of John Wiley and Sons, Inc.
Fig. 4.7 Equation 4.18 taken from H. Ito, Trans. ASME 82, Series D, 121
Fig. 8.2 E. N. Seider and G. E. Tate, Ind. Eng. Chem. 28, 1429 (1936). 245
123-34 (1959).
Reprinted by permission of the American Chemical Society.
Fig. 4.8 Steam-Its Generation and Use, Babcock and Wilcox, New York, 122
Fig. 8.3 W. Nunner, VDI-Forschungsheft No. 455/1956 (VDI-Verlag 246
1963, pages 8-17. GmbH-Dusseldorf).
Fig. 4.13 J. H. Perry, Chemical Engineers' Handbook, third edition, McGraw- 129
Fig. 8.4 R. A. Seban and T. T. Shimazaki, Trans. ASME13, 803 (1951). 246
Hill, New York, 1950, pages 405-406. Used with permission of
Fig. 8.5 T. K. Sherwood and R. L. Pigford, Absorption and Extraction, 250
McGraw-Hill Book Company.
second edition, McGraw-Hill, New York, 1952, page 70; F. Eisner,
Fig. 4.14 J. H. Perry, Chemical Engineers' Handbook, third edition, McGraw- 130
Proc. 3rd Intern. Congr. Appl. Mech., 1930, page 32.
Hill, New York, 1950, pages 405-406. Used with permission of Fig. 8.6 W. E. Ranzand W.R. Marshall, Jr., Chem. Eng. Prog. 48, 141-146, 250
McGraw-Hill Book Company.
173-180 (1952).
Figure-Table Source Page Figure-Table Source Page
Fig. 8.7 R. A. Seban and D. L. Doughty, Trans. ASME78, 217 (1956). 251 Fig. 11.9 G. G. Gubareff, J.E. Janssen, and R.H. Torborg, Thermal Radia- 369
Fig. 8.8 W. H. McAdams, Heat Transmissions, third edition, McGraw-Hill, 253 tion Properties Survey, Honeywell Research Center, Minneapolis,
New York, 1954, pages 173-176. Used with permission ofMcGraw- 1960.
Hill Book Company. Figs. T. J. Love, Radiative Heat Transfer, Merrill, Columbus, Ohio, 376-378
Fig. 8.10 W. M. Rohsenow, Developments in Heat Transfer, MIT Press, 257 11.14-11.16 1968.
Cambridge, Massachusetts, 1964, Chapter 8. Fig. 11.21 H. C. Hottel and A. F. Sarofim, Radiative Transfer, McGraw-Hill, 384
Table 8.1 M. A. Grossman, Elements of Hardenability, ASM, Cleveland, 259 New York, 1967. Used with permission of McGraw-Hill Book
1952. Company.
Fig. 8.12 Reproduced by permission of Progressive Metallurgical Industries, 260 Fig. 11.22 H. C. Hottel and A. F. Sarofim, Radiative Transfer, McGraw-Hill, 385
Gardena, California. New York, 1967. Used with permission of McGraw-Hill Book
Fig. 8.13 G. Stolz, V. Paschkis, C. F. Bonilla, and G. Acevedo, JISI 193, 261 Company.
116-123 (1959). Fig. 11.23 F. Kreith, Principles ofHeat Transfer, second edition, International 388
Fig. 8.14 V. Paschkis and G. Stolz, Heat and Mass Flow Analyzer Labora- 261 Textbook Co., 1965, page 231.
tory, Columbia University. Fig. 11.25 Adapted from H. C. Hottel and A. F. Sarofim, Radiative Transfer, 391
Fig. 8.15 Adapted from P. M .. Auman, D. K. Griffiths, and D.R. Hill, Iron 262 McGraw-Hill, New York, 1967, Chapter 6. Used with permission
and Steel Engineer, Sept., 1967. of McGraw-Hill Book Company.
Fig. 8.16 N. V. Zimin, UPC 621.784.06, pages 854-858, Plenum Press, New 263 Fig. 11.26 Adapted from H. C. Hottel and A. F. Sarofim,.Radiative Transfer, 391
York, 1968. Translated from Metallovedenie i Termicheskaya McGraw-Hill, New York, 1967, Chapter 6. Used with permission
Obrabotka Metallov, No. 11, 62-68 (Nov., 1967). of McGraw-Hill Book Company.
Fig. 8.18 E. M. Sparrow and R. D. Cess, J. of Heat Transfer, Trans. ASME, 268 Fig. 11.27 Adapted from H. C. Hottel and A. F. Sarofim, Radiative Transfer, 392
149-156 (May, 1962). McGraw-Hill, New York, 1967, Chapter 6. Used with permission
Fig. 8.19 P. Berenson, J. Heat Transfer, ASME, Aug., 1961, page 35. 269 of McGraw-Hill Book Company.
Table 8.2 W. M. Rohsenow, Developments in Heat Transfer, Chapter 8, MIT, 271 Fig. 11.28 F. Kreith, Principles ofHeat Transfer, second edition, International 392
Cambridge, Massachusetts, 1964. Textbook Co., 1965, page 231.
Fig. 9.5 P. J. Schneider, Conduction Heat Transfer, Addison-Wesley, 284 Fig. 11.29 Adapted from H. C. Hottel and A. F. Sarofim, Radiative Transfer, 393
Reading, Massachusetts, 1955, page29. McGraw-Hill, New York, 1967, Chapter 6. Used with permission
Fig. 9.16 V. Paschkis, Welding Research Supplement, Sept., 1946, pages 310 of McGraw-Hill Book Company.
497-502. By permission of American Welding Society. Fig. 11.30 Adapted from H. C. Hottel and A. F. Sarofim, Radiative Transfer, 393
Fig. 9.17 V. Paschkis, Welding Research Supplement, Sept., 1946, pages 312 McGraw-Hill, New York, 1967, Chapter 6. Used with permission
497-502. By permission of American Welding Society. of McGraw-Hill Book Company.
Fig. 9.23 P. J. Schneider, Conduction Heat Transfer, Addison-Wesley, 320 Table 11.2 H. C. Hottel, "Radiation,~' Chapter IV of Heat Transmission, by 394
Reading, Massachusetts, 1955, page 266. W. H. McAdams, McGraw-Hill, 1954. Used with permission of
Fig. 10.7 C. M. Adams, "Thermal Considerations in Freezing,"' Liquid 339 McGraw-Hill Book Company.
Metals and Solidification, ASM, Cleveland, Ohio, 1958. Fig. 11.31 P. J. Schneider, Temperature Response Charts, Wiley, New York, 399
Fig. 10.8 C. M. Adams, "Thermal Considerations in Freezing," Liquid 340 1963. Reprinted by permission of John Wiley and Sons, Inc.
Metals and Solidification, ASM, Cleveland, Ohio, 1958. Fig. 12.5 Adapted from B. I. Kitaev, Y. G. Yaroshenko, and V. D. Suchkov, 413
Fig. 10.10 C. C. Reynolds, Trans. American Foundrymen's Society 17, 343 342 Heat Exchange in Shaft Furnaces, Pergamon Press, New York,
(1964). 1968. Reprinted with permission.
Figs. A. W. D. Hills and M. R. Moore, Heat and Mass Transfer in 355 Fig. 12.11 J. F. Elliott, Trans. AIME 227, 806 (1963). By permission of The 421
10.16-10.18 Process Metallurgy, The Institution of Mines and Metallurgy, Metallurgical Society of AIME.
London, 1967. Fig. 12.14 Adapted from B. I. Kitaev, Y. G. Yaroshenko, and V. D. Suchkov, 422
Fig. 11.8 , G. G. Gubareff, J.E. Janssen, and R.H. Torborg, Thermal Radia- 367 Heat Exchange in Shaft Furnaces, Pergamon Press, New York,
tion Properties Survey, Honeywell Research Center, Minneapolis, 1968. Reprinted with permission.
1960. Fig. 12.15 R. A. Limons and H. M. Kraner, paper in Agglomeration, Inter- 424
Table 11.1 H. C. Hottel and A. F. Sarofim, Radiative Transfer, McGraw-Hill, 368 Science-Wiley, 1962. Reprinted by permission of John Wiley and
New York, 1967. Used with permission of McGraw-Hill Book Sons, Inc.
Company. Fig. 12.16 C. G. Thomas, article in Energy Management in Iron and Steelworks, 426
Figure-Table Source Page Figure-Table Source Page
Special Publication No. 105, Iron and Steel Inst., London, 1968, Table 16.1 P. Clausing, Annalen der Physik 12, 976 (1932). 561
page 87. Fig. 16.9 Adapted from D. L. Schroeder and J. F. Elliott, Trans. AIME 236, 563
Fig. 13.4 J.E. Reynolds, B. L. Averbach, and M. Cohen, Acta Met. 5, 29 435 1091 (1967). By permission ofThe Metallurgical Society of AIME.
(1957). Reprinted with permission of Pergamon Press. Fig. 16.10 R. G. Ward, J. Iron and Steel Inst. 201, 11 (1963). 566
Figs. O. D. Sherby and M. T. Simnad, Trans. ASM 54, 227 (1961). 435-436 Table 16.2 R. Ohno and T. Ishida, J. Iron and Steel Inst., London 206, 904 567
13.5-13.6 (1968).
Table 13.1 K. Compaan and Y. Haven, Trans. Faraday Soc. 52, 786 (1956). 437 Table 11.1 Adapted from North American Combustion Handbook, published 573
Fig. 13.13 D. Mapother, H. N. Crooks, and R. Mauer, J. Chem. Phys. 18, 452 by North American Manufacturing Co., Cleveland, Ohio 44105,
1231 (1950). 1952.
Tables L. Yang and G. Derge, paper in Physical Chemistry of Process 460,462 Table 111.1 Adapted from North American Combustion Handbook, published 586
13.4-13.5 Metallurgy, AIME7, 195 (1961). By permission ofThe Metallurgi- by North American Manufacturing Co., Cleveland, Ohio 44105',
cal Society of AIME. 1952.
Table 13.7 E. T. Turkdogan, paper in Steelmaking, The Chipman Conference, 466 Tables J. F. Elliott et al., Thermochemistry for Steelmaking, Vol. 2, 588-598
J. F. Elliott (ed.), Addison-Wesley, Reading, Massachusetts, 1962. lV.l-IV.10 Addison-Wesley Publishing Co., Reading, Massachusetts, 1963.
Table 13.8 E. N. Fuller, P. D. Schettler, and J. C. Giddings, Indust. and Eng. 467
Chem. 58, 19 (May, 1966). Reprinted by permission of the Ameri-
can Chemical Society.
Fig. 13.20 Adapted from L. N. Satterfield and T. K. Sherwood, The Role of 469
Diffusion in Catalysis, Addison-Wesley, Reading, Massachusetts,
1963.
Table 13.9 From L N. Satterfield and T. K. Sherwood, The Role of Diffusion 471
in Catalysis, Addison-Wesley, Reading, Massachusetts, 1963.
Fig. 14.1 R. P. Smith, Acta Met. 1, 578 (1953). Reprinted with permission of 474
Pergamon Press.
Fig. N. L. Peterson and R. F. Ogilvie, Trans. AIME218, 439 (1960). By 485
Example 14.3 permission of The Metallurgical Society of AIME ..
Fig. 14.14 D.R. Porier, R. V. Barone, H. D. Brody, and M. C. Flemings, J. 498
Iron and Steel Inst., 371 (April, 1970).
Fig. 15.4 R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phe- 529
nomena, Wiley, New York, 1960, page 664. Reprinted by per-
mission of John Wiley and Sons, Inc.
Figs. J.P. Hartnett and E. R. G. Eckert, Trans. ASME 79, 247 (1957). 536-537
15.6-15.7
Fig. 15.8 Data from E. D. Taylor, L. Burris, and C. J. Geankoplis, I. and E. C. 539
Fundamentals 4, 119 (1965), reprinted by permission of the Ameri-
can Chemical Society; and T. R. Johnson, R. D. Pierce, and W. J.
Walsh, ANL Report on Contract W-31-109-eng-38, 1966.
Fig. 15.9 R. G. Olsson, V. Koump, and T. F. Perzak, Trans. AIME236, 426 541
(1965), reprinted by permission of The Metallurgical Society of
AIME; 0. Angeles, G. H. Geiger, and C.R. Loper, Trans. Ameri-
can Foundrymen's Society 76, 629 (1968); and M. Kosaka and S.
Minowa, Trans. Japan Iron and Steel Inst. 8, 393 (1968).
Fig. 16.4 W. M. McKewan, Trans. AIME 212, 791 (1958). By permission of 557
The Metallurgical Society of AIME.
Fig. 16.5 E. Kawasaki, J. Sanscrainte, and T. J. Walsh, AIChEJ8, 48 (1962). 558
Fig. 16.7 R. G. Olsson and W. M. McKewan, Trans. AIME236, 1518 (1966). 559
By permission of The Metallurgical Society of AIME.
PART ONE
FLUID DYNAMICS
The first part of this text deals with fluids, their intrinsic properties, their behavior
under various conditions, and the methods by which we can manipulate and utilize
them to produce desired results. Most metallurgical processes deal with fluids at
one point or another, and although the metallurgist is usually not required to be an
"expert" on fluids, he should understand the fundamentals of fluid dynamics as
presented in the following chapters, and be able to make intelligent use of the
properties offtuids and characteristics of equipment used to manipulate and control
fluids.
The behavior of fluids is also intimately related to heat and mass transport
processes. For example, if a gas is hotter than a solid past which it is flowing, the
solid naturally is heated. The rate at which heat is transferred to the solid 's surface
is dependent on the fluid's properties and its flow pattern. Similarly, if a piece of
graphite is to be dissolved in a bath of molten iron, the rate of dissolution depends
on the motion of the liquid iron adjacent to the graP,hite. Thus, to appreciate
transfer of heat and/or mass, some understanding of fluid dynamics is important.
These are just two simple examples which illustrate.the necessity for the student
to become acquainted with the means of examining and expressing the flow of
ft uids and to eventually recognize the role of fluid flow in rate processes involving
heat and mass transfer. Actually, if we take a fundamental approach to the study
of fluid flow, then the subject matter is appropriately designated momentum
transport. Momentum transport with energy transport and mass transport make
up the subject of transport phenomena, which Bird, Stewart, and Lightfoot* rank as
a "key engineering science," along with thenpodynamics and mechanics. Trans-
port phenomena as a key engineering science has been well received in both the
academic and industrial communities of engineering.
* R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, Wiley, New York, 1960.
1
PROPERTIES OF FLUIDS
~~
For laminar flow the velocity profile is parabolic; in turbulent flow, the curve is
somewhat flattened in the middle.
F/A
1J = YI= v;y'
Units of 17 are: 2
F/A = 0.05 lb J 32.2 lbm ft/sec _ lbm
2 61
lb ft- 2 ft 1 lb J - 1. ft sec 2
1J = (fthr~1)(ft-1) = lbfhrft-2 Also
Alternatively the English system yields the following units for 17: V 5 ft/sec
= - 1-f- = 480sec- 1.
_ (lbmfthr- 2)(ft- 2) _ _1 _1 Y %t
1]- (fthr-l)(fCl) -lbmhr ft. Then
In the cgs system of units, the poise (P) is used, in which 1.61 lbm 3600 sec _ lb
1J = = 1208-m
1 poise (P) = 1 dyn sec cm - 2. 480 ft sec 1 hr ft hr
The centipoise (cP) is probably the most common unit tabulated for viscosity. It or
equals 0.01 poise, and is the viscosity of water at 68.4F. Thus the value of the 12.08
viscosity in centipoises is an indication of the viscosity of the fluid relative to that 1J = .4 cP = 4.99 cP.
2 2
of water at 68.4F.
In many problems involving viscosity, it is useful to have a value of a fluid's
viscosity divided by its density p. Hence we define the kinematic viscosity v at this 1.3 VISCOSITY OF GASES.
point as For the purpose of explaining momentum transport in gases, we resort to the
simplest tre.atment of the kinetic theory of gases. We utilize the concept of the
v = !1.. mean free path, in which the molecules are idealized as billiard balls, and postulate
p
a hypothetical "ideal" gas possessing the following features:
The kinematic viscosity is a lundamental quantity, in that it is a measure of
momentum diffusivity, analogous to thermal and mass diffusivities, which will be 1. The molecules are .hard spheres resembling billiard balls, having diameter d
and mass m.
2. The molecules exert no force on one another except when they collide. stated above, the number of molecules arriving from above and below y 1 is equal,
3. The collisions are perfectly elastic and obey the classical laws of conservation and on the ave~age no ~et ~omentum is transferred across the plane y 1 .
of momentum and energy. To det~rmme ~he v1scos1ty of the gas, consider the gas under the influence of
~ac~oscop1c flow m the x-direction, with a velocity gradient dvxfdy, as depicted
4. The molecules are uniformly distributed in a concentration of n per unit m Fig. 1.5.
volume throughout the gas. They are in a state of continuous motion and are
separated by distances which are large compared to their diameter. Velocity profile v,(y)
5. All directions of molecular velocities are equally probable. The speed (magni-
tude of velocity) of a molecule can have any value between zero and infinity.
If we assume that the molecules possess a Maxwellian speed distribution (i.e., y = Y1
the thermal energy of the gas is given by the total kinetic energy of all the moving
molecules), then the average speed Vis given by
V=V8K8 T (1.4)
nm x
where K 8 is the Boltzmann constant and Tis absolute temperature. Fig. .1.5 Relation between the macroscopic velocity profile and the plane of interest at y
1
In addition,, for such a collection of molecules, a significant parameter that
governs the mechanism of momentum transfer in gases is the free path, defined as No~ if all the molecules have velocities characteristic of the plane in which they
the distance traveled by a molecule between two successive collisions. At the last collided, we may write the x-momentum above y 1 as
instant of collision, the center-to-center distance of two molecules is d. Intuitively
we know that the mean free path A. should be inversely proportional to the collision
mvx/ _ = mvx/ (1.8)
cross section nd 2 , and also inversely proportional to the concentration n of the YI +y YI
molecules. The rigorous analysis for determining A. includes these terms, along
Similarly, for below y 1
with a coefficient whose numerical value is developed by considering the random
fluctuations of the colliding molecules. The final result gives
mvx/ - = mvx/ - ~Am dvx. (1.9)
A.= (Ji}(n:2J (1.5) Y1-y YI dy
We find the net rate of x-momentum crossing the plane y 1 by summing the
Now consider an imaginary plane at y = Yi. which is being crossed by molecules x-momentum of molecules that cross from below and subtracting the x-momentum
in either direction. If we examine the condition of no bulk motion (no macroscopic of those that cross from above. In this manner, we write
flow) of the gas in they-direction, then the molecules cross the y 1 plane from both
sides with equal frequency. This frequency per unit area at whic,h molecules cross ryx = zm[vxl _ - vxl ]. (l.10)
the plane at y 1 from one side is given by Y1-y Y1+y
good agreement with experimental data up to about ten atmospheres. However, teristic diameter of the molecule in AJ(see Fig. 1.6). The qU:an ty n~ is the collision
the temperature dependency is only qualitatively correct, in that Y/ does increase --integral. of the Chapman-Enskog~tneory, which is a function of a dimensionless
with increasing temperatures; but, quantitatively, the temperature dependency is temperature parameter 1< 8 T/s. In order to use Eq. (1.14), we need values of (J and
not satisfactory. Data for real gase_s indicate that Y/ varies with rn, with n between s/1< 8 ; these parameters are known for many substances. A partial list of them is
0.6 and unity, rather than 0.5, as in Eq. (1.13). given in Table 1.1. We can then determine the collision integral by using Table 1.2.
The more up-to-date kinetic theories replace the billiard-ball model with a
more realistic molecular force field by considering the force of attraction and Table 1.1 Intermolecular force parameters and critical properties
repulsion between molecules. These theories, reviewed by Hirschfelder, Curtiss,
and Bird, 1 make use of the potential energy of interaction between a pair of Lennard-Jones
molecules in the gas. This function-often referred to as the Lennard-Jones parameters* Critical constantst
potential-d*>plays the behavior of molecular interactions by exhibiting weak Molecular '
attraction at large separation distances and strong rermlsion at small separations, Substance weight,M t a,A E/KB, K '!;'.,OK V0 cm 3 /g-mol
as shown in Fig. 1.6. '---
Light elements
H2 2.016 2.915 38.0 33.3 65.0
He 4.003 2.576 10.2 5.26 57.8
Noble gases
Ne 20.183 2.789 35.7 44.5 41.7
yRepulsion Ar 39.944 3.418 124. 151. 75.2
Kr 83.80 3.498 225. 209.4 92.2
Attraction Xe 131.3- 4.055 229. 289.8 118.8
Simple polyatomic
substances
~ir 28.97 3.617 97.0 132. 86.6
N2. 28.()2 3.681 91.5 126.2 90.1
J~2 32.00 3.433 113. 154.4 74.4
co 28.01 3.590 110. 133. 93.l
-:C02 44.01 3.996 190: 304.2 94.0
S0 2 64.07 4.290 252. 430.7 122.
Fig. 1.6 Lennard-Jones potential function describing the interaction of two nonpolar F2 38.00 3.653 112.
molecules. .-c1 2 70.91 4.1J5 357. 417. 124.
Br 2 159.83 4.268 520: 584. 144.
The equilibrium position of the molecules is located at 15, where the potential CH 4 16.04 3.822 137. 190.7 99.3
energy is a minimum at -s; sis called the characteristic energy parameter. Using
{>
* 1. 0.Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids, Wiley, New
York, 1954.
1
J. O. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids, t K. A. Kobe and R. E. Lynn, Jr., Chem. Rev. 52, 117-236 (1952), and Amer. Petroleum Inst. Research
Wil~y, New York, 1954. Proj. 44, edited by F. D. Rossini, Carnegie Inst. of Technology, 1952.
~
a are not available, one may use a modified form of Eq. (l.14), as presented by
Table 1.2 Values of 0-integral for vis- Bromley and Wilke 2 :
cosity and of the viscosity function
f(KBT/s), based on the Lennard-Jones
potential* Y/ = 3.33 x 10
_ 5 JAiTc (KBT)
v,.213 . f 7 . (1.15)
+ -Mi)-112[ 1 + (Y/;)112(Mi)1'4]2
6.0 0.8963 2.751 _ 1 (
".{ /fI . 8.0 0.8538 3.337 <l>ij - -;;; 1 - - .
3.866
...; 8 Mi Y/i M;
10 0.8242
r1 -:.--.-io 0.7432 .4,... 6.063 Here n is the number of chemical species in the mixture; X; and xi are the mole
'.lJ4o 0.6718 ' - 9.488 fractions of species i and j; Y/i and Y/ i are the viscosities of species i and j ,at the system
60 0.6335 12.324 temperature and pressure; and M; and Mi are the corresponding molecular weights.
80 0.6076 14.839
Note that <!>ii is dimensionless, and, when i = j, <l>ii = 1.
100 0.5882 17.137
To summarize, Eqs. (1.14), (1.15), and (l.f6) are useful equations for computing
200 0.5320 26.80
41.90
viscosities ofnonpolar gases and gas mixtur~ at low density from tabulated values
400 0.4811
of the intermolecular force parameters a and s. They cannot, however, be applied
* J. 0. Hirschfelder, C. F. Curtiss, and R. B. with confidence to gases consisting of polar or highly elol)gated molecules such as
Bird, Molecular Theory of Gases and Liquids, H 2 0, NH 3 , CH 3 0H, and NOC!. A further limitation is that for the most part
Wiley, New York, 1954. these equations have been tested only over the temperature range 100K.to 1500K.
T~e data on viscosity of several gases as a function of temperature are given in
Example 1.2 Compute the viscosity of hydrogen at 1 atm and 2000F (1364K)j Fig. 1. 7. Keep in mind that the data are valid for pressures up to about 10 atmos-
pheres. Note that (1) the viscosity of all gases increases with temperature, and (2)
Solution. From Table 1.1, we find that the magnitude of viscosity does not solely depend on the molecular weight of the
s/K8 = 38.0 K, a= 2.915A. gas. For ex~mple, the data for helium fall in the range of miwh heavier gases than
hydrogen. Sports commentators in particular should heed some of the data and
From Table 1.2, stop propagating the myth that baseballs can be hit farther in dry air than under
Q~ ~ 0.69. humid weather conditions.
Sub'stituting appropriate values into Eq. (1.14), we have
1.4 VISCOSITY OF LIQUIDS
- 2.67 x 10:5 (2j(.9152)()1364
YJ - . )
) -- 248 x
2(
0 69
. 10-
4
poise. In dealing with transport processes in liquids, we are always faced with the problem
2
(Observed viscosity is 2.44 x 10- 4 poise.) L. R. Bromley and C.R. Wilke, Ind. Eng. Chem. 43, 1641 (1951).
3
C.R. Wilke, J. Chem. Phys.18, 517-519 (1950).
Using Eq. (1.14) to calculate Y/ requires knowing~ and s/K8-. When values of.
. Se~eral theor!es have been postulated to account for th~ observed properties
that much less is known about the structure of liquids than about the structures of
ofhqmds, none without some serious difficulties. The oldest, in terms of the struc-
solids or gases. However, there is more similarity hetween liquids and solids than
tural picture involved, is the hole theory, which postulates that a liquid contains
between liquids and gases. This similarity is based on the small fractional increase
many holes or vacant areas, distributed throughout the liquid, with these holes
in volume on melting (3 to 5 ~;;'., for metals), and the fact that the heat of fusion is
having some distribution of sizes about an average. Although this theory does not
quite a bit less than the heat of vaporization. Furthermore, x-ray data tell us that
explain all observations of the changes of properties of materials upon fusion, it has
there is at least some degree of short-range order in a liquid. That is, at a short
been useful in deriving a relatively simple approach to predicting the temperature
distance from a central atom, the arrangement of nearest neighbors is reasonably
dependence of the viscosity of liquids. Other models, based on differing assump-
predictable. However, as the distance increases, the predictability of atom positions
tions of where the extra volume associated with a liquid is assumed to reside, will
decreases rapidly, unlike in solids.
be discussed in Chapter 13.
'l:;
Because liquids near their melting points still represent a dense phase, the
0.06 -~ concept of transfer of momentum from atom to atom, as utilized in the kinetic
~ theory approach to gas-phase viscosity, is invalid, since the momentum of each
~
> atom varies rapidly with the vibration of the atoms within the pseudo-lattice of the
liquid. As Frenkel 4 points out, the fact to be explained in the case of liquids is not
their r~sistance to shearing stress, but rather their capability of yielding to stress.
0.05
Without recourse to any specific model of a liquid, but assuming only that
viscous flow takes place by movement of particles past other particles, we can start
by considering the mobility of an individual particle. Einstein has shown that the
mobility B of a particle under the influence of an external force is related to the
0.04 diffusion coefficient D by the relationship
(1.17)
Bis the mean velocity divided by the force acting on the particle. Since diffusion
appears to be an activated process, i.e., a minimum activation energy 11Gt must be
0.03
supplied to the particle to move it from one stable position to the next then the
~uidity, which is proportional to the capability of atoms to move, just a; diffusion
1s, must also be thermally activated. <-
0.01
11 = 11G;is] ,
A exp [ RT (1.18)
{
80
J70
24,
1-100 50% NaOH, 50% H 2 0 J60
70 22
Linseed oil JSO
60 J40 20
0.1-1.0 H 2 S0 4 J30
50 J20 J8
y
Molten salts 40
JJO
J6
l
JOO
Heavy metals (Pb, Au, Zn, etc.)
90 J4
Alkaline earth metals (Ca, Mg) 30
80
O.ol-0.1 Transition met.als (Fe, Ni, Co, etc.) J2
70
20
Water (70F) 60 JO
Kerosene (70F) JO so
40
{ Acetone 30
JO
-20
Figure 1.8 gives a nomograph for the viscosity of common liquids. More -JO 8 10 12 14 16 18 20
specific aspects of sev('ral classes of liquids of particular interest to metallurgists -JO -20
x O.J
Ca * 17fJ2No
Y/ = JMRT.' (1.22)
T* = K8 T, (1.23)
.
bs
and
1
V*=-
.. (1.24)
n[J3
0.2
The variables used are:
2000K 1000K 5009K 300K
0.15...__+----+--~-~--~--~~~-
0.5 1.0 1.5 2.o 2.5 3.0 3.5 (J= interatomic distance in the close-packed crystal at 0K, A,
Temperature-, K-1. s = energy parameter characteristic of specific metal,
Fig. 1.9 The viscosities of liquid metals and their dependence on temperature. (From N 0 = Avogadro's number.
Chapman, ibid.) M = atomic weight,
R = gas constant,
Chapman 5 has analyzed these data in the light of the theory of liquids devel- T = absolute temperature, K,
oped by Kirkwood 6 and by Born and Green. 7 He arrived at a generalized model Ks = Boltzmann's constant, and
for the viscosity of liquid metals which involves no assumptions of the structure n = number of atoms per unit vol.ume.
other than that the atoms are spherical and that the potential between atoms can J-c0
be expressed by some function </J(r) of the distance between atoms and an energy The parameter (J is taken as the interatomic spacing for the close-packed
parameter, such as is done in the Lennard-Jones picture of the potential energy crystal at 0K. The energy parameters present the largest difficulty, and have been
well between atoms, in which derived in the following maimer. The effective Lennard-Jones parameters for
sodium and potassium have been determined. 8 Using these two values, one can
plot the reduced viscosity data for sodium and potassium as a function of reduced
temperature. The data points fall on a smooth curve, as predicted by Eq. (1.21).
5
Then, assuming that all the rest of the metals in Fig. 1.9 obey the same functional
T. Chapman, AIChE Journal 12, 395 (1966).
6
relationship, the viscosity-temperature data for !he remainder of the pure metals
J. G. KirI<wood, J. Chem. Phys. 14, 180 (1946).
7
M. Born and H. S. Green, A General Kinetic Theory of Liquids, University Press, Cambridge,
8
1949. R. C. Ling, J. Chem. Phys. 25, 609 (1956).
are correlated by empirically adjusting the parameter s/Ka until all the data for a The fact that, by adjusting an unmeasured parameter, it is possible to correlate
given metal fall on one point on the curve, given in Fig. 1.10. Table 1.4 shows the all these liquid-metal viscosities on a single curve, might not be taken as significant,
resulting values of s/Ka starting with the known (measured) values for sodium and except for the fact that one theory of melting indicates that the melting point should
potassium. be proportional to s, and that this relationship has been observed for other classes
of materials. The empirically determined values of s/Ka from Table 1.4 have been
plotted as a function of the absolute melting temperature, and an excellent correla-
tion has been observed, leading to the equation
f, - 0
- - 5.20Tmelting K. (1.25)
Ka .
For metals with high melting points, this may be used to predict viscosities, as no
further data are available.
Example 1.3 Estimate the viscosity of liquid titanium at 1850C. The following
data are available: Tm = 1800<;::. M = 47.9.g/mol,- sp. gr. = 4.50 g/cm 3 , and
3 4 6 7
<> = 2.89A.
l/T*
Fig. 1.10 Correlation' curve for viscosities of liqu.id metals. (From Chapman, ibid.) Solution. Using Eq. (1.25), one can estimate s/Ka to be
s/Ka = (5.20)(2073) = 10,780K.
' '
Table 1.4 Empirically determined values Then T* is give'n by Eq. (1.23), apd the product 11*(V*) 2 is found from Fig. 1.10:
of a/K8 (from Chapman, ibid.)
T* 1
Metals b,A e/1's, oK = 10,780 (2123) = 0.~97,
a+(3
5000 20
5 10 15
Wt% silicon
Fig. 1.11 Viscosities of aluminum-silicon alloys. (From W.R. D. Jones and W. L. Bartlett, ,
J. Inst. Metals 81, 145 (1952).) Si-0 tetrahedra
u..
0
~- 1600 _
~ i::-------------
~ 1500 }7'~""'===::::::::::--------.:
~
2600
2400
1300
1200 2200
1100 2000.
1000 1800
('y + Cem.)
0.0 , l.O 2.0 3.0 4.0 5.0 \ I
Wt% carbon Crystalline silica Molten silica
(a) (b)
Fig. 1.12'- Viscosities of iron-carbon alloys. (From R. N. Barfield and J. A. Kitchener, J. Iron
and Steel Inst. 180, 324 (1955).) Fig. 1.13 Silicate tetrahedron and the structure of silica.
or Si 4 + ions, and the structure built up is a regular crystalline array of SiO!- Table 1.5 Structural relationships in basic oxide-silicate melts
groups. This is illustrated in Fig. l.13(a). When this substance is melted, the
arrangement presumably continues, but the long-range order is destroyed, as Total oxygen atoms Corresponding binary
Structure Equivalent silicate ion
indicated in Fig. 1.13(b). However, the same Si-0 bonds are present and these Silicon atoms molecular formula
high energy bonds need to be broken so that viscous flow could take place. The
activation energy for this process is quite high (135 kcal) and the viscosity of pure 2: 1 Si0 2 All corners of Infinite network
liquid Si0 2 at 1940C is 1.5 x 105 poises. tetrahedra shared
When CaO, or another similar divalent basic oxide, is added to the melt, the 5:2 M02Si0 2 One broken link (Si 6 0 15 ) 6 - or
Ca 2 + ions are accommodated in the interstices of the silicate structure and the per tetrahedron (Sis020) 8 -
0 2 - ions enter into the network (Fig. 1.14). The 0 2 - ions cause two of the tetra-
3:1 MOSi0 2 Two broken links (Si 3 0 9 ) 6 - or
hedra to separate since each corner of these two can now have an oxygen ion of
per tetrahedron (fing) (Si 4 0n) 8 -
its own. Increasing additions of base results in a progressive breakdown of the
6
original three-dimensional network. 7:2 3M02Si0 2 Three broken links (Si 2 0 7) -
per tetrahedron (chain)
Ca2+
4:1 2MO Si0 2 All links broken Discrete (Si0 4 ) 4 -
Si 4 +
(orthosilicate) tetrahedra
easier as the size of the anions decreases (Fig. 1.15), so that the activation energy
.!l~is falls off continuously as basic oxide is added (Fig. 1.16), up to the orthosilicate
I I
0 0 composition, beyond which no further significant decrease is expected. At that
I I
-O-Si-0-Si-O- +Cao=
point the network of (Si0 4 ) 4 - tetrahedra is completely broken down. For com-
I I positions more basic than the orthosilicate, the viscosity changes that do occur
0 0
I I are more likely due to changes in the liquidus temperature.
I I 0 140
0 0 __ Si0 2
I I Li
'""
0
-O-Si-0- + -0-Si-0- +Cai+
-"' 120 o Na
I I
0 0 ++~ ;::,. K
I I \.:>
<1 Mg
\
JOO \ Ca
\ Sr
\ & Ba
80 \
Fig. 1.14 The solution of a divalent metal oxide in molten silica. \
\
Table 1.5 shows this progression as a function of total oxygen atoms to silicon
atoms. As the breakdown progresses, we may consider the silicate network to be
60 \\
made up of p.rogressively smaller discrete ions. From the original three-dimensional
network, we progress to discrete molecules, such as (Si 6 0 15 ) 6 - and (Si 3 0 9 ) 6 - , as
more and more oxygen ions are contributed by the basic oxide. Actually, it is more
probable that rather than only one type of silicate ion existing at any given ratio
40
20
' ':
D ~..:...,.J!
of base to acid, there exists a statistical distribution of sizes about the average one, 10 20 30
as indicated by the number of links broken per tetrahedron. Mo!% rfietal oxide
As the'{hree-dimensional network breaks down, the number of Si-0 bonds
Fig. 1.15 Decrease of the activation energy of viscosity for silica as basic oxides are added.
that need to be broken during viscous flow decreases. The shear process becomes (From Bills, ibid. 201.)
~~-~iJJIBiS to .molt'..J.i:_'!cJions, anc:Likt~e-..Xa- from Fig. 1.17. Magnesium
-~
0. oxide (MgO) is ~g~iv~Lent to CaO, u_p_lQ.. '!9out 10 mole percenL~gQ, and theiL.,
"" 0.8 ~ions~.are adde(ffogetlier- to obtain Xe~~. Thus, when we add Xs;o 2
.Q
and Xcv_ and know the temJ!!llilture we can find the viscosit fro ig. 1.18. Tliis
~eliavior ~f~} 2 ~jsJl!rther:..ilh1_s_!:rnt~<!_~y ig. _1._l_~ \Vhi~h s_h9_~s the viscosit~
0.4 of CaO~i~.slags_(!t_J50Q C. Note th<ttJh_ej_ov_iscosity lines are essentiall~
0
...l 4.4
Si0 2 35-70
MgO 0-25
Mo! fraction CaO
4.0
Fig. 1.16 Viscosity of the SiOi-CaO solutions. (From J. 0. M. Bockris, J. A. Kitchener, and
J. Mackenzie, Trans. Faraday Soc. 51, 1734 (1955).) 3.6
Alumina, which appears to exist in molten oxide solutions as (Al0 3 ) 3 - anions, 3.2
behaves in much the same way as silica. However, silica and alumina are not 2.8
equivalent on a molar basis, since the basic building block of alumina is (Al0 3 ) 3 -,
and two Al3+ ions can replace two Si 4 + ions only if one Ca 2+ ion is available to 2.4
maintain electrical neutrality. Thus, as far as its effect on viscosity is concerned, 2.0
alumina has a silica equivalence Xa, which depends on the Al 20 3 /Ca0 ratio and
on the total Al 20 3 content, as shown in Fig. 1.17. ~kd.:.~!!..g_ 1.6
these data in correlatingjhe viscosity and composition o1Ca0-Mg0-Al 20 3 -Si0 2 l.2
slags, have found that, fo;agjyen temer~ure,_a_smQQ.th curve_c_orrelates all
c'Omposrtion-Sin the studied rang;(Fig~ 1js). To utilize this fact, we conve!t the- 0.8
0.4-
0 0.6
.S'
0
-0.4
0.4 0.5 0.6 0.7 x~io, 0.8
0.4
Xs;o~+ Xa
lime molar ratios. (From Bills, ibid. 201.) Finally, we. shou.Id note the ability of the well-known flux CaF 2 (ftuorspar) to
10
9
E.T. Turkdogan and P. M. Bills, Amer. Ceramic Soc. Bull. 39, 682 (1960). P. M. Bills, J. Iron and Steel Inst. 201, 133 (1963).
f
(
decrease the viscosity of oxide slags. Although we do not quite understand the
reason, it may be due to the ability of CaF+ cations to move between silicate
anions that are electrostatically bound together, by mutual attraction to a Ca 2 +
cation, thereby decreasing the electrostatic bond and thus lowering the viscosity.
The effect is much more pronounced at low rather than at elevated temperatures,
and in acidic slags more than in basic slags. Figures 1.21 and 1.22 illustrate the
effect for basic and acidic slags, respectively.
Temperature, F
2500 2600 2700 2800 2900 3000
.?& 15
0 4%CaF 2 2% CaF 2 0% CaF 2
0.
~
;;i
>~ 10
2%MgF 2
\
\
\
\
"----->L-L_-'L-----''--"-~::...._:.t,-----"'--~---"~-~~----"'---~30 \
20
Wt%Ca0
70
',
O~~_,_...__~~_,_~-..,..--~~--...,.-,,'~--....,-,.,..,.._--.,-:c:'.
Fig. 1.19 Isoviscosity lines (poise) in the Ca0-Al 2 0 3 -Si0 2 system at 1500C. (From J. S. 1350 1400 1450 1500 1550 1600 1650
Machin and T. B. Yee, J. Am. Cer. Soc. 31, 200 (1948).) Temperature, C
Fig. 1.21 Influence of CaF 2 and MgF 2 additions on the viscosity of a basic slag. Slag com-
position: Ca0-51.7%, Mg0-3.2%, Al 2 0 3 -12.7%, Si0 2 -32.4%. (From Elliott, ibid.)
- ..!.- Valid for 1250C
~ Si0 2 -CaO-MgO- I
"i3 2.6
0. Al 2 0 3 systems I l 300C 1:- 2.6
-~(3
;;i
2.4 containing I / 0
0-16%Fe0 / / ~ 2.4
~
-;;:
;;: 2.2 x Slags containing I / // OJ)
0 2.2
OJ)
0
16-30% FeO I / ...l
...l 2.0 I I I
1
1.8 1 I I 1350C
/ I I
/ I I
1.6
/ /I
1.4 , x' ,x' / /
/ /
1.2
,x )( /
/ _,x /
/X /X
1.0 /X/
/ x/ x
x_,, xx
0.8 .,,... 0 8
'"'12~00~-,~3""0~0--1-4~0-0--15~0-o-
x
x Temperature, C
0.6
x
Fig. 1.22 Influence of CaF 2 on viscosity of 50 % Si0 2 -l 5 % Al 2 0 3 -35 % CaO melt (acid
0.4 0.5 0.6 0.7 slag). (From Bills, ibid. 201.)
XsiOi = 0.390, _x.A1io, = 0.069. . Accordin? to Ne~ton's law of viscosity (Eq. 1.3) the shear stress 'Yx' plotted versus
For our purposes, we lump Xcao and XMgO together, so the effective Xcao = the velocity g~ad1ent -dvxfdy, should yield a straight line running through the
0.539. ongm. Expenmen_tal~y, ~his has been proved true for all gases and for all single-
Calculating the ratio XA 120)Xcao (=0.128), enter Fig. 1.17 at XA1 20 , = 0.069, phase nonpolymenc hqmds. Fluids that behave in this manner-and most fluids
move slightly to the right of X Alio,/ X cao 2 = 0.2, and go down to find X a '.c::'. 0.10. do -are termed Newtonian fluids. However, Eq. (1.3) does not describe the be-
After calculatingXsioi + X 0 = 0.490, move to Fig. 1.18, and go up to the c~rves havior of a large number of fluids which are called non-Newtonian fluids and
for 1600C and 1500C. The viscosity values are 1.8 and 2.9 poises, respectively. include substances such as molten plastics, slurries, and certain slags.
The study of Newtonian flow is part of a larger discipline of science known as
The simplest molten salts are those in which we take the bonding to. be
rheolog~. Rh~ology encompasses the mechanical. behavior of gases, liquids, and
essentially electrostatic; for example, we consider that molten N_aCl cons~sts
solids, mciudmg Newtonian gases and liquids on the one hand and Hookean
entirely of Na+ and c1- ions. _Since there appe~~ t? be _no large am~n or cat10n
behavior of solids on the other. The spectrum of substances between these two
complexes analogous to the (S10 4 ) 4 - or (Al0 3 ) . 10n_s m molten ox1~e syste~s,
extremes concerns the engineers and scientists dealing with non-Newtonian fluids.
and since there is a relatively large fraction of nondlfect10nal electrostatic ?ond_1~g,
A large number of non-Newtonian fluid follows one of the behavioral
flow takes place relatively easily in these materials. Figure 1.23 shows v1scos1t1es
patterns represented in Fig. 1.24, where the shear stress r is plotted versus the
of many molten salts. Note that their viscosities are about 1/lOOth of those of the
rate_ofstrai~ Y(y equals -dvxfdy at steady state). In contr;st, the curve ofstress-
molten oxides. stram rate is also plotted for Newtonian fluids, demonstrating the obvious fact
Because of the nondirectionality of the bond forces, the activation energies of that ryx/Y is independent ofY. On the other hand, non-Newtonian fluids are those
viscosity for salts are also lower than those of the slags. in which ryx/Y is a function ofY.
T,oc
300 100
1200 800 600 400
"""u 50.0 .c>""
_;:. ~"
;;;
/
0 30.0
:;l
> 20.0
::<,b
10.0 ~
,;;;"""'
7.0
I (,'>
CY'
~uCI
5.0
3.0
2.0
1.0
0.7
0.5
0.3 20 25
5 10 15
_!_ oK-1 x 10
T Strain rate, i'
Fig. 1.23 Viscosities of molten salts. (From C. J. Smithells, Metals Reference Book, fourth
Fig. 1.24 Stress-strain rate curves for time-independent fluids.
edition, Vol. I, Plenum Press, New York, 1967.)
The fluids classified as Bingham plastics require a finite shear stress T 0 (yield
stress) to initiate flow. In other words, the fluid remains rigid when the shear removed and we can see that the GNF moves no further while the viscoelastic
stress is less than T 0 , but flows when the shear stress exceeds T 0 . An example of a fluid goes backward. It does not, however, return to its original position, recoiling
only partly.
flu"id exhibiting Bingham plastic behavior is an aqueous slurry of fine, powdered,
coal. The following relationship gives the rate of shear stress and shear strain:
Tyx = To+ YJpy, Tyx > To. (1.26)
I Charcoal line
(a)
Here YJP is a plastic viscosity or coefficient of rigidity. We use the plus sign when y
is positive, and the minus sign when y is negative.
Pseudoplastic fluids, which are characterized by a decreasing slope of the Tyx D (b)
D
versus y curve as the stress increases, need, like Newtonian fluids, no yield stress
~
for flow, but, unlike in Newtonian fluids, Tyx/Y does depend on Tyx Dilatant fluids
differ from pseudoplastics in that the Trx versus y curve has an increasing slope as
(c)
! j
stress increases. For both pseudoplastic and dilatant fluids we can sometimes use
~
~
the Ostwald power law to describe the relationship of the stress-strain rate: (d)
11
~
R. B. Bird, Can. J. qfChem. Eng. 43, 161(Aug.1965).
Stationary
1.2 Water at 100F flows over a flat plate. Si0 2 We can observe even more dramatic effects when oxides such as Li 2 0 and Na 2 0 are
a) If the velocity profile at a point x = x 1 is given by vx = 3y - y 3 , find the shear stress at added to silica; for example, for 0.5 mole fraction Si02 , the activation energy is only 23 kcaljmol.
the wall at that point. Explain with the aid of sketches, why this is so.
a) 100,
b) 10,
c) 1.
What do the results mean to you?
1"6' What is the viscosity of chromium at 2000C? Data are as follows:
Melting point, 1898C
Atomic wt, 52.01
Specific gravity, 7.1
. fJ,2.nA
:;, 'ti}i.using Bills' metho~,estimate. the :iscosity ofa 50 wt% Si0 2 , 30 wt% CaO, 20 wt% Al 2 0 3
sfag at 1500C. Does this agree with Fig. 1.19?
1.8 At 1000C a melt of Cu-40 % Zn has a viscosity 5 cP; at 950C the same alloy has a vis-
cosity of~ cP. 'Estim~te the viscosity of the alloy at l l00C.
1.9 When oxides such as MgO and CaO are added to molten silica, the activation energy of
viscosity is reduced from 135 kcal/mo! for pure Si0 2 to about 39 kcal/mo! for 0.5 mole fraction
Momen~um in (?rout) ~ay e~ter a system by momentum transfer according to
Newton s equat10n of v1scos1ty (if the fluid is Newtonian; otherwise various
equations for non-Newtonian fluids are used), or it may enter due to the overall
2 fluid motion. The forces applied to the balance are pressure forces and/or gravity
forces.
The momentum balance is actually a force balance because we are concerned
LAMINAR FLOW AND MOMENTUM with the rate of momentum that enters and leaves the unit volume. Units of momen-
BALANCES tum are M L/T (M = mass, L = length, T = time), whereas a rate of momentum is
2
M L/T Classical physics states clearly that forces (F = ma) are involved when
we consider momentum rates. Thus, if the term momentum balance confuses the
reader, he is reassured that a force balance is being applied.
In discussing Newton's law of viscosity, we have described fluid motion as flowing
parallel layers which, due to viscosity, establish a velocity gradient dependent upon 2.2 FLOW OF A FALLING FILM
the shear stress applied to the fluid. This velocity gradient has been regarded as a ~on.sid.er the flow of a liquid at steady state along an inclined plane (Fig. 2.1). The
potential or a "reason" for momentum transport from layer to layer. hqmd is at a constant temperature, and therefore its density and viscosity are
In this chapter, we shall first derive simple differential equations of momentum constant. Furthermore, we consider only that portion of the plane where the
for special cases of flow, for example, flow of a falling film, flow between two entrance and exit of the liquid to the plane are sufficiently removed so as not to
parallel plates, and flow through tubes. To make it possible for the student to influence the velocity vz. In this situation, vz is not a function of z but obviously a
participate in developing hairy formulas, these derivations make use of the function of x. * Figure 2.1 also depicts the unit volume as a "shell" with a thickness
concept of a momentum balance and the definition of viscosity. These classic dx and length L ; the width of the shell extends a distance W, perpendicular to the
examples of viscous flow patterns certainly apply to rather simplified and idealized page. The terms used in Eq. (2.1) are as follows.
conditions. You may be tempted, therefore, to disregard the importance of
thoroughly understanding these examples; however, we should point out that, Rate of momentum in
despite the simplicity of the following calculations, you will gain an appreciation across surface at x
of the variables involved. Also, you will obtain a basic tool for analyzing engineer- (moment transport due to viscosity)
ing problems: the ability to arrive at pertinent differential equations. Rate of momentum out
across surface at x + dx
2.1 MOMENTUM BALANCE* (due to viscosity)
A momentum balance is applied to a small control volume of fluid to develop a Rate of momentum in
differential equation. The differential equations, when their solutions comply with across surface at z = 0
the physical restrictions (boundary conditions), yield the algebraic relationship (due to fluid motion)
which can be used to determine the engineering characteristics of the system. The
solutions give the velocity distributions from which other characteristics, including Rate of momentum out
the shear stress at the fluid-solid interface, are developed. As we shall see in across surface at z = L
Chapter 3, the shear stress at the fluid-solid interface is very important in analyzing (due to fluid motion)
L) ''
Gravity force acting on fluid _,.~ ,
the disposition of energy flowing through a system. For steady state flow, the (LWdx)(pg cos {3)
momentum balance is
In this particular problem, the pressure forces are irrelevant because the
rate of ) ( rate of ) ( sum of forces ) pressure is equal thr?ughout the liqui.d. Also note that all terms in the list, including
( momentum in - momentum out (2.1)
+ acting on system = O. the first t"'.o, ar~ z-directed forces. Figure 2.1 shows that momentum in by viscous
transport IS X-dlfeCted, but if we think of interpreting rxz in an alternative way-
*The general aspects of the developments in Sections 2.1-2.6 are similar to those found_ in
Chapters 2-3 in R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, Wiley,
New York, 1960. * In the region where vz = J(x) and vz =l= f(z), we say that the flow is fully developed.
36
namely, as a shear stress-we certainly realize that we are dealing with a z-directed We have now recognized the definition of the first derivative of 'xz with respect to
x, and have thus developed the differential equation pertinent to our system:
force.
drxz
Entrance disturbance dx = pg cos (J. (2.4)
Liquid film l
l This equation is integrated to yield
'xz = pgx COS fJ + C 1 . (2.5)
Equation (2.5) describes the momentum flux (or alternatively the shear-stress
Exit distribution), but contains an integration constant C1 . This constant is evaluated
by recognizing that the shear stress in the liquid is very nearly zero at a liquid-gas
(a)
Momentum interface. In other words, the gas phase, in this instance, offers little resistance to
in by flow liquid flow, which results in a realistic boundary condition:
at x = 0, Lxz ,;; q. (2.6)
Substitution of this boundary condition into Eq. (2:5) requires that C 1 = O; hence
the momentum flux is
r xz = pgx. cos (J. (2. 7)
If the fluid is Newtonian, then we know that the momentum flux is related to
velocity gradient according to
(2.8)
Direction
of gravity
Substituting this expression for 'xz in Eq. (2.7) give~ the distribution of the velocity~
(Momentum\
\
(b)
gradient:
out by flow \ pg cos /3.
x. (2.9)
\ 11
x=O x=l
(2.22)
~
3
Q= W b (Po - PL).
(2.23)
3 17 L
On looking back through this example, we note that in this instance the fluid
flows because of the pressure drop (P0 - Pd. For horizontal flow, such a pressure
Fig. 2.2 Flow between parallel plates. drop would be necessary to make the fluid flow, in contrast to the flow down
an i~clined plane (Section 2.2) on which gravity exerts the necessary force for fluid
motion.
Again, the momentum in and out of the system due to fluid motion are equal. We
are left with
/
(LW)(Tyx)ly - (LW)(Tyx)lyHy + cPo - PL) dyW = 0. (2.16) C/2.4 FLOW THROUGH A CIRCULAR TUBE
Dividing through by L W dy and letting dy approach zero, we develop the differen- Conside_r the fully developed flow of a fluid in a long tube of length Land radius R;
tial equation we specify fully developed flow so that end effects are negligible. Since we are
dealmg with a pipe, it is convenient to work with cylindrical coordinates. Therefore
d(Tyx) cPo - Pd the shell in- Fig. 2.3 is cylindrical, of thickness Ar and length L.
(2.17)
dy L
Rate of momentum in
The"~oundary conditions are described at the centerline (y = 0) and at the solid across surface at r
(2nrfa,,)I,
wall ( y = <5) as follows: (due to viscosity)
B.C. 1 at y = 0, Note that here we include the area factor (21!rL) in parentheses. This is because
(2.18)
B.C. 2 at y = b, the area as well as the shear stress is a function of r.
Momentum in by flow Pressure P 0 At r = o, the velocity gradient (hence, the shear stress) equals zero; this can be
realized because of the symmetry of flow.
Thus for this case,
at r = O,._ (2.27)
B.C. 1
< ~
and noting
B.C. 2 at r = R, Vz = 0, (2.30)
Gravity force actihg on the cylindrical shell (2nr f'lrL)pg , , ii) The average velocity is
Pressure force acting op surface at z = 0
Pressure force acti~g on surface at z = Ll
(2nr Ar)Po'
-(2nr Ar)PL
1
~~
11
~V 1 (p _:_ P'-L +pg)R8'1).
27tR
v.J!~ -J Jv r.dr-de = o
2
,
(2.33)
.~"-~~,' nR 2
z - L YJ
. \ 0 0 ,. -
We now add up the contributions to the momentUJll balance: '-........_:
;) - \
iii) Volume flow rate is
(2nrL;'L"rz)lr - (2nrfarz)lr+&r + 3'1!_r ~rf:ipg + 2nr:'Kf(P0 - PL) = 0. (2.24)
Q =(Po~ PL+ pg) (n~ )
4
Note that all terms contain the factor r.; however, sin~e r is a variable, it shoulo (2.34)
n()t be used as a common divisor. By dividing through by 'i:dt f'lr arid takiu'g the
limit as f'lr goes to zer,o, we develop the differential equation This latter result which is commonly referred to as the Hagen-Poiseuille law,
is valid for laminar' steady-state flow of incompressible fluids in tubes having
d (nrz) (Po -L PL sufficient length to make end effects negligible. An entrance length give? by Le.=
dr = + pg )"r. (2.25)
0.035 DRe is required before we can establish fully developed parabohc velocity
distribution.
Integration. yields
Example 2.2 Water at 60F flows through a horizontal tube of diameter 1/16 in
("\ (P0 -PL
+ pg )r- + -
C with a pressure drop of 0.04 psi/ft. Find the mass flow rate through the tube.
1
'L"rz.)H'
L 2 r \ r
r.: ;.:.~ ~~
Solution. In this situation, the force of gravity does not act on the fluid in the 2.5 {ENEltAL M0MENTUM EQUATIONS
direction of flow, so according to Eq. (2.34) the volume flow rate is
In the previous sections of this chapter, we determined velocity distributions for .
Q =(Po - PL)nR 4
_ some simple flow systems by applying differential momentum balances. The
L 811 balances for these systems served to illustrate the application of the principle of
Substituting in values, we obtain the following terms: conservation of momentum. In general, when dealing with isothermal fluid
systems which do not involve changes in compositions, we can solve problems by
0.04lbf 144in 2 starting with general expressions. This method is better than developing formula-
f . 2
t Ill f 2
t = 5. 76 lbJ/ft 3 ' tions peculiar to the specific problem at hand. "fhe general momentun:i b.~Jance ~
5 called the equation of motion or the Navier-Stokes' equation; in addition the
_ l.14cPl2.09 x 10- lbfsec/ft 2 _ _ 5 lb 2
equation of continuity is frequently used ill conjunction with ihe-eq~jt~fmo!i~if
1J - cP - 238
. x 10 J sec/ft ,
1 The continuity equation is developed simply by applyillg the law of conserva-
1 1 tion of mass to a small volume element within a flowing fluid. The principle of
R = (2)(16)(12) ft = 384 ft. conservation of mass is quite simple to apply and we assume that the reader has
used it in developing material balances. We develop the equation of motion by
Therefore applying the law of conservation of momentum which, in its general form, is an
extension of Eq. (2.1 ). With the aid of these two equations, we can mathematically
Q= (5.76)(i)( 3 ~ 4 )4( 2 _ 38 : 10 _ 5) == 4.31 6 3
x 10- ft /sec. describe the problems enc;ountered in the previous section, as well as more compli-
cated problems. However, as we shall see, these expressions are rather cumbersome,
Thus we see that the mass flow rate and exact solutions can be found only in very limited cases. Hence these equations
are used primarily as starting points for solving problems. The equations of
pQ = (62.4 lbm/ft 3 )(4.31 x 10- 6 ft 3 /sec) continuity and motion are simplified to fit the problem at hand. Although theoreti-
= 2.74 x 10- 4 lbm/sec, cally these equations are valid for both laminar and turbulent flows, in practice
or they are applied only to laminar flow.
pQ = 0.980 lbm/h.
We should then check if the flow is laminar, by evaluating the Reynolds number. 2.5.tvk'quation of continuity
As mentioned in Chapter 1, the criterion is Re < 2100: Consider the stationary volume element within a fluid moving with a velocity
DV DVp ,, having the components vx, vY, and v,, as shown in Fig. 2.4. We begin with the
Re=-=-- basic representation of the conservation of mass:
v 1J
Also rate of mass ) ( rate of) ( rate of )
( accumulation = mass in - mass t 2 35
- pQ ( )
pV = nD 2 /4'
First, look at the faces perpendicular to the x-axis. The volume flow rate of
so that the Reynolds number may be written in the alternative form fluid in across the face at xis simply the product of the velocity (x-component) and
the cross-sectional area, yielding dy dzvxlx The rate of mass in through the face
Re= DpQ 4pQ at x is then dy dz(pvx)lx. Similarly, the rate of mass out through the face at x + dx
(nD 2 /4)1J = nD11' is dy dz(pvx)lxHx We may write analogous expressions for the other two pairs
(4)(62.4)(4.31 x 10- 6 x 3600) of faces, and then enter all the terms that account for the fluid entering and leaving
'.i Re (consistent units)
, must be used the system into the mass balance, and leave the accumulation term to be developed.
(n)(l/(l2 x 16))(1.14 x 2.42)
The accumulation is the rate of change of mass within the control volume
" = 86.
ap
dxdydz-
Since Re < 2100, the flow is laminar. at
(x + ti.x.y + ti.y,z + ti.z)
I
I Txx +ax
y -------"'\_
x x
Fig. 2.4 Volume element fixed in space with fluid flowing through it. (b)
(a)
Fig. 2.5 Momentum transport (x-component) due to viscosity into the volume element.
The mass balance thus becomes (a) Directions of viscous momentum transport. (b) Directions of forces.
ap
Ax Ay Az at = Ay Az[pvxlx - PVxlx+&-"'] Figure 2.5(a) shows the x-components of -r as if they were made up of viscous
momentum fluxes rather than shear stresses. On the other hand Fig. 2.5(b) shows
+ Ax Az[pvYIY - pvyly+&y] the x-components of -r as stresses. Note the appearance of <xx which by the scheme
+ Ax Ay[pvzlz - PVzlz+&zJ. (2.36) of subscripts represents the transport of x-momentum in the x-direction. Alterna-
tively, we view <xx as the x-directed normal stress on the x-face, in contrast to
Then, dividing through by Ax Ay Liz, and taking the limit as these dimensions <yx which we view as the x-directed shear stress on they-face.
approach zero, we get the equation of continuity: Let us now develop the terms that enter into Eq. (2.39). First, the net rate at
Normal avy 2
stresses
Tyy = -217 ay + 317(V v) (2.50) The bracketed terms on the left side of these equations merit attention.
Consider a control volume of fluid moving in space with no mass flow across its
+ t17
avz surface. The change in the x-component of its velocity with time and position is
'!zz = - 217 Tz 3 (V v) (2.51)
given by
= = -1](0Vz + OVx)
. Avx . {ovx At ovx
a= hm-= hm - - + - - + - - + - - '
Ax
ovx Ay ovx Az} (2.60)
T
zx
T
xz ox oz (2.54) x 81 _ 0 At dt-O Ot At OX At Oy At OZ At
These equations constitute a more general statement of Newton's law of viscosity we obtain
than that given in Eq. (1.2), and apply to complex flow situations. When the fluid
flows between two parallel plates in the x-direction so that vx is a function of y o~ o~ o~ o~ n~
alone, where they-direction is perpendicular to the plates' surfaces (Fig. 1.4), then
ax=-+
fu 'x~
-+ vzfu
v--+ vY~ -=-
lli
(2.61)
In terms ofr:,
OVx OVx OV~ OVx) i' P
x-component p (- +v - +v - +v - = --
ot x OX y oy 'z oz i!x
CT xx OTyx
- (- + - +OTzx)
- +pgx (A)
Table 2.1 The continuity.equation in different coordinate systems* OX oy az
DP
*Tables 2.1-2.7 are from R. B. Bird, W. E. Stewart, and E. ~'.Lightfoot, Transport Phenomena, Wiley,
oy
New York, 1960, pages 83-91. Reprinted by permission.
_,.-,.,-------~--------------------------------------, . . . .- - - - - - - - - - - - - - - - - - - - - - . . . , . - . . _ . . - ' V ' . . , . .. --'-~UQ~...,..1-n-~u1-,.-11111~'1-~1nl"l"UIU'Zl'1.~~7
Table 2.4' The conservation of momentum in spherical coordinates (r, IJ, </>)
In terms of r:
+ _I_.
r sm e o</> r (A)
1a 1 OT,e Tee OT,z)
- (--(rr ) + - - - - + - +pg, (A)
r or rr r ae r oz
IJ-component OVe OVe Ve OVe v</> OVe v,ve v~ cote)
p (- + v - + - - + - - - + - - - - -
ot r or r ae r sin e o</> r r
,.,,, IJ-component
1 oP (1 a 2 1 a 1 ore"'
- - - - --(rr,e)+-.--(reesinlJ)+-.--
r ae r2 or r sm e ae . r sm e o</>
(B)
T,e COt IJ )
+ -r - -r- '"'"' + pge (B)
z-component
(C)
In terms of velocity gradients for a Newtonian fluid with constant p and 11:
- -oP
or
+ 11 ( \7 2Vr - -r22 Vr - -r22 -OVe
ae
2
- - 2 Ve COt IJ
r
2 av"') (D)
- r2 sin e o</> + pg,
1. "
I
ot r or r ae r sin e o<f>
-
r r
2 cos e OVe)
+ r2 sin 2 e o</> + pg</> (F)
/
Table 2.5 Components of the stress tensor in rectangular coordinates (x, y, z)
Table 2. 7 Components of the stress tensor in spherical
coordinates (r, e; </>)
<xx= -11 [ 2 ovx
OX - 2
3(Vv) J (A)
rrr = av, -
-11 [ 2a,:- 2
3(Vv) J (A)
rYY = -11 [ 2OVy
Oy- -
2
3(Vv) J (B)
too= ilv
2 ( -;:1 Je
0
+-;:v,) - 2
3(V -.;v) J (B)
zz = -11 [ 28z
Ovz - 2
t(V v) J (C)
-I] [
(C)
rxy = <yx = -11 [ovx
ay + avyJ
ax (D)
L (D)
!yz = !zy = -11 [avy
a;+ avz]
ay (E)
sin() a ( vq, ) 1 ilv 0] (E)
[avz avx] roq, = rq,o = - I] [ -r- iJ() sin () + r sin 8 il<f> c:~./-::;":",
rzx = xz = - '1 ax + a; (F) =
(F) ~-----i
(Vv)=-+-+-
aux avy avz (G)
,D ,-;
L 't"zr = r,z = - I]
[av~ av,]
ar + Bz \ (F) Fig. 2.6 Velocity profile and momentum boundary layer of flow parallel to a flat plate.
velocity V00 At any point x downstream from the leading edge of the plate, we 1 * vx * x b* = ~
observe that the velocity increases from zero at the ~all to very near V at a very
00
l u = -v- ' x = I,' L'
00
short distance b from the plate. The loci of positions where vxfV = 0.99, is b, and
00
it is defined as the boundary layer. At the leading edge of the plate (x = 0), b is y
y* = _,
zero, and it progressively grows as flow proceeds down the plate. L
Whenever problems of this type are encountered, namely, in the flow of fluid
past a stationary solid, the viscous effects are felt only within the fluid near the solid,
vz
FrL = ~
that is, y < b. Of course, this is exactly where the behavior of the fluid should be gYL
analyzed, because for y > b, the happenings from the point of view of this discussion Substituting these parameters into Eqs. (2.64), (2.65), and the continuity equation,
are essentially uneventful, due to the fact that in this region vx is essentially uniform we get
and constant, being equal to V00 Since vx is uniform and constant for y > b, Eq. (A)
J Continuity:
in Table- 2.2 reveals that the pressure gradient oP/ox is zero or, stated differently,,
pressure everywhere in the bulk stream is uniform. In turn, the pressure within
the boundary layer is equal to the pressure in the bulk stream, so that oP/ox is also (2.66)
zero within the boundary layer. We are now almost ready to pick out the appro-
priate equation of motion for the flow pattern in Fig. 2.6, but before we do so, let Momentum:
us examine which velocity components are relevant.
Ii -~ As discussed above, vx is a function of y, and the determination of this functional (2.67)
i [ l/ relationship is indeed a major part of describing the flow and how the fluid and
solid surfaces interact. Also note that vx depends on x. This results from the fact
that, as the fluid progresses down the plate, it is retarded more and more by the (2.68)
drag at the plate's surface. Thus ovx/ox is not zero, and the equation of continuity
for steady two-dimensional flow of fluid with constant p and '1 is The next step is to make order-of-magnitude estimates of the terms in Eqs. (2.67)
ovx OVy and (2.68), realizing that we are interested only in the happenings within the
0 (2.63) boundary layer.
ox + oy =
Order of magnitudes:
Thus Vy exists and we should consider both the x- and y-components in Table 2.2.
For the steady-state case with constant density and viscosity, Eqs. (D) and (E) dx* ~ 0 to 1, (2.69)
in Table 2.2 reduce to
dy* ~ b*, ";:l (2.70)
du*~ I.
L
(2.64) (2. 71)
From (2.69) and (2.71), ~e get
and
2 2 (2.72)
OVy OVy (o vy o vy)
Vx OX + Vy oy = V OX2 + oy2 + gy. (2.65)
and from continuity,
When we remember that we are primarily interested in the region y :=; b, at this
point it is convenient to define some dimensionless parameters :t ov*
oy* ~ I. (2.73)
t The Rsynolds number ReL which was briefly introduced in Chl!,pter I reappears here again. From (2.70) and (2.71), we obtain
Also the Froude number FrL is introduced. These two dimensionless groups of variables,
which so often occur in engineering studies, have been given names in honor of those two early
workers in fluid mechanics. (2.74)
-~-----u~I:7RHl"llTUl____.l~nr-IIIUIIl'l;IlUtll-ln:ll'illl\;~--------------------;..-,-------,,
and from (2.72) and (2.73), A similarity argumenti shows that the stream function may be expressed ~s
(2.75)
I/! =~ f(/3), (2.85)
where
P~ ~f~
Now we c"n estimate all the second derivatives:
(2.86)
~,~1~ (2.76) From Eqs. (2.83), (2.85), and (2.86),
ol/! ol/! ofJ df
Vx = oy = 0{3 oy = Voo d/J' (2.87)
(2.77)
o v*
2
,.,_, ()* (2.78)
v = _ ol/! =
y OX
~.2...;-;-
gvV (P d[J
00
df -V)~
(2.88)
o(x*) 2 = ,
Then Eq. (2.84) becomes
o v*
2
,.,_, 1
(2.79)
o(y*)2 = l>*.
(2.89)
Insertion of the various magnitudes into Eqs. (2.67) and (2.68) reveals two
important facts: o 2u*/o(y*) 2 o 2u*/o(x*) 2, and the equation involving the Mathematically, the use ofljJ and f3 has reduced a partial differential equation to an
x-component of the velocity has much larger terms than that for vY. Hence we ordinary differential equation with the boundary conditions also taking equivalent
deal only with forms:
ovx OVx o 2vx df
Vx OX + Vy oy = V oy2 , (2.80) B.C. 1 at f3 = 0, f = 0, df3 = O; (2.90)
which is the boundary layer equation for a flat plate with zero pressure gradient. df
We now proceed to solve Eq. (2.80) for the boundary conditions B.C. 2 at f3 = oo, d[J = 1. (2.91)
B.C. 1 at y = 0, Vy= O; (2.81) Equation (2.89) may be solved by expressing f(/3) in a power: series, that is,
00 '
B.C. 2 at y = oo, (2.82)
f =I ak[Jk. The technique is too involved to develop here, but the solution
0
In order to simplify Eq. (2.80), we define the stream function I/! ast conforming to the boundary conditions becomes
ol/! ol/! a/3 2 1 a2 f3 5 11 a 3 /J 8
vx -= -oy and vy -= - -
ox (2.83)
!=2!-25!+ 48!+ ... , (2.92)
The use of the stream functions simplifies Eq. (2.80) and automatically satisfies wher~ a = 0.332. Then Eqs. (2.87) and (2.88) give expressions for vx and vY; the
continuity (Eq. 2.63). Substituting Eq. (2.83) into Eq. (2.80) yields: ~olution for vx is shown graphically in Fig. 2.7. The position,~0.,99,
is locate~ a!_J3__:::5.D-;-thas-t-he-buurrdary-layer-thickness hr"
ol/! 0 21/! ol/! 0 21/!
* ()
3
- -- - - - -2 = v0-1/!3 (2.84)
= 5.0;~-.
oy ox oy ax ay oy
(2.93)
_t The stream function is not discussed in this text, but it is an important tool often used in fluid
dynamics. A discussion of the stream function can be found in V. L. Streeter, Fluid Dynamics,
McGraw-Hill, New York, 1948, pages 38-41.
---
~Blasius showed that Eq. (2.84) could be solved in this manner.
em., 1949, page 1217.)
.;.
(2.95) x --+--____,_____,___.----1---_
'"'----,:lx------1...j
x
In other words, the shear stress at the solid surface is integrated over the entire x+.:lx
surface. From Fig. 2.7 we find that
Fig. 2.8 Integrated mass and momentum quantities in the boundary layer over a flat plate.
o(vx/Voo)J ~ 0.332. (2.96)
[ ap /3=0 For unit thickness of fluid, the mass flow rate at x is
Knowing the integrand in Eq. (2.95), we can now perform the integration. The
result is (2.98)
(2.97)
and the momentum flow is
This is the drag force exerted by the fluid on one surface only.
1.0
Mx = J PVxVx dy. (2.99)
0
~1~ The net change in the mass flow rate between x and dx is
II Q5 h
Then in order to satisfy continuity, an amount of fluid d Wx must enter the control
element through its top surface, y = h.t The momentum associated with this
fluid is therefore '
Fig. 2.7 Solution for the velocity distribution in the boundary layer over a fiat plate. (From
L. Howarth, Proc. Roy. Soc., London Al64, 547 (1938).) (2.101)
We can now write a momentum balance that also includes the viscous momentum
2.7.2 Flow over a flat plate-approximate integral method leaving at y = O:
The exact solution method yields the velocity profile, boundary layer, and the drag
force on the plate. In this section, we give an alternative solution which provides
only the boundary layer thickness and drag force. Consider the portion of the
developing boundary layer between x and x + dx in Fig. 2.8.
--
Lh;,
'
(2.102)
t The reader should satisfy himself that fluid does enter rather than leave the control element
top'"'"~
~~----------------~---------~------~------ -- - - - - - - - - - - - - - - - - - - --~----------------r-myy----------~~--~---------,--
d
dx [fh PVx
2
dy Jo/</ - Voo dxd [fh pvxdyJ /= -11 (av~)
~lX
U)' y=O /
/
~k
Equations (2.107) and (2.108) compare very favorably with the exact results
given by Eqs. (2.94) and (2.97), respectively. This is just one example ~here an
0 I o I integral technique is found to be useful. It is also a powerful tool for. solvmg many
heat transfer problems in which exact solutions are not always available. There-
Both integrals are split into two, that is,
fore later in the text, we will discuss this method again.
' -r. .v::1-'~
h b
2.7.3 Flow in inl~t of circular tubes
J=J+f,
0 0 b
In Section 2.4, we considered the fl.ow of fluid in a long tube so that end effects
were negligible. Now we wish to study the fl.ow conditions at the inlet where the
and when we realize that between J and h, vx = V00 , then Eq. (2.102) reduces to fl.ow is not fully_de_yelope_d_. The fluid enters the tube with a uniform velocity V0
in the ~-dir~~tion. -The important component for this system is the z-component,
just as the x-component is most important for the fl.ow past a fl.at plate. According
(2.103)
to Eq. (F) in Table 2.3, the momentum equation with ~=_9, avzf at~= 9, and
gz = o,_ reduces to _f' ,,
for constant p and 17.
Equation (2.103) is known as the van Karman integral relation for zero pressure v 8vz
r ar
+ v avz
z az
= -~ 8P + v[~ ~(r
p az r ar
8vz)
ar
+ 8vz]
az
2
2
gradient. It is solved by assuming a reasonable velocity distribution. Even with
,~I an assumed distribution that departs from the actual distribution, the results for Again, the viscous effect in the direction parallel to fl.ow is negligible, so that
t I the drag force and boundary-layer thickness are very good. 82 vzf}z 2 ~ 0, and we are left with
I, The velocity distribution is assumed to have the following form:
11i
1.0
When this is substituted into Eq. (2.103) and the indicated operations are per~
r vz
formed, a differential equation for the boundary layer result~: ~
R. IOOR21'
0
for t~is problem, ~s descri~ed in ~ig. 2.9. His analysis shows that a fully developed
flow is ?ot established ~nhl vz/~ V0 ~. 0.07. Th~s an entrance length (z = Le) of The continuity equation (Eq. (C) in Table 2.1) is
approximately 0.035 (D Vop)/11 is reqmred for bmldup to the parabolic profile of 1 a 2 1 a . e) = 0 .
the fully developed flow. . 2 -;-(r v,)
r ur
+ -.-() ~e (v Sill
r Sill u 9 (2.113)
2. 7.4 Creeping flow around a solid sphere The momentum flux-distribution, pressure distribution, and velocity com-
ponents have been found analytically: 3
C~nsider the flow of an incompressible fluid about a solid sphere (Fig. 2.1 O). The
~ 11 V ~\
4
flmd approaches the sphere upward along the z-axis with a uniform velocity v r -
00
sin(), (2.114)
-2 R r]
9 (
00 '
Radius of sphere "R z
R (R)
2
~ P = P0 -
3 11V00
pgz - 2. -;:- cos (), (2.115)
At every point there are
pressure and friction
~( ~) H~) 3J
forces acting on the
sphere surface
(x,y.z) v, = voo [ 1 - + cos(), (2.116)
-f or
, / 1 (r.e,q,)
,/' I
~(!!_) - ~(!!_)
3
,/' I
I v9 = - V00 [ 1 - ] sin () (2.117)
I 4r 4r
I
I
I
'------i--1- - - - y
We can check the validity of the results by showing that Eqs. (2.111)-(2.113)
I and the following conditions are satisfied:
x ---.... I
Projection
B.C. 1 at r = R,
of point on
xy-plane
B.C. 2 at r = oo,
In Eq. (2.115), the quantity P0 is the pressure in the plane z = 0 far away from
the sphere, - pgz is simply the hydrostatic effect, and the term containing V00
t Fluid approaches
from below with
~'" velocity v~
results from the fluid flow around the sphere. These equations are valid for a
Reynolds number (DV00 /v) less than approximately unity.
With these results, we can calculate the net force which is exerted by the fluid
Fig. 2.10 Coordinate system used in describing the flow of a fluid about a rigid sphere. on the sphere. This force is computed by integrating the normal force and tangential
force over the sphere surface as follows.
Clearly, the mo~ent~m equatio_n for this situation does not involve the -com-
The normal force acting on the solid surface is due to the pressure given by
ponent. In add1t10n, 1f the flow 1s slow enough, the acceleration terms in Navier- Eq. (2.115) with r = Rand z = R cos e. Thus
Stokes' equation can be ignored. Therefore, in spherical coordinates, from
Eqs. (E) and (F) in Table 2.4, we obtain for the r-component: 311Voo
P (r = R ) = P0 - pgR cos () - - - - cos ().
2 R
oP ( 2 2 2 ov 2 ) The z-component of this pressure multiplied by the surface area on which it
- Or + 11 V V, - r2 V, - r2 O; - r2 V9 C.Ot () + pg, = 0, (2.111)
2
acts, R sin() d() d</J, is integrated over the surface of the sphere to yield the net
force due to the pressure difference:
and for the e-component:
2it "
Equation (2.118), integrated, yields two terms: Solution. A force balance on the sphere, as it falls through the liquid, is made
according to the diagram:
(2.119)
F; = ff
0 0
(T,i=R sin 8)R 2 sin 8 d8 dc/J.
Here Fs is the buoyant force exerted by the liquid and is therefore directed upward;
FK is often called the dragforce and as such always acts in the opposite direction to
so that the friction drag results: that of motion and is therefore directed upward. The only force in the downward
direction is the weight of the sphere. The terminal velocity is reached when the
(2.120)
force system is in equilibrium. Therefore Fs + FK = Fw, and by substituting in
The total force F of the fluid on the sphere is given by the sum of Eqs. (2.119) and expressions for each of these forces, we have
(2.120):
(2.124)
where Ps is the sphere's density.
(2.121)
By solving Eq. (2.124:) for 17, we arrive at
These two terms are designated as F., (the force exerted even if the fluid is stationary) 2R 2 (Ps - p)g
and FK (the force associated wi'th fluid moveme~t). Thus these forces are '1 = (2.125)
9Vt
t
, This result is' valid only if 2R Vr/v is less than approximately unity.
(2.122)'
I PROBLEMS
F K = 6nlf/}R V00 :::;fR 1ec tOH (2.123)
2.1 A continuous sheet of metal is cold-rolled by passing vertically between rolls. Before , , f
We use Eq. (2.121), known as Stokes' law, primarily for determining the terminal I entering the rolls, the sheet passes through a tank of lubricating oil equipped with a squeegeeP:~~
velocity, VP of small spherical particles moving through fluid media. The fluid ..! device that coats both sides of the sheet uniformly as it exits. The amount of oil that is carried
media are stagnant; the spherical particle moves through the fluid, and V is through can be controlled by adjusting the squeegee device. Prepare a control chart that can
00 be used to determine the thickness of oil (in inches) on the plate just before it enters the roll
viewed as Vi. With this in mind, we may use Stokes' law as the basis of a falling-
as a function of the volume r~te of oil (in lbm per hour). Values of interest for the thickness of
sphere viscometer, in which the liquid is placed in a tall transparent cylinder and a ~ the oil film range from 0-0.024 in.
sphere of known mass and diameter is dropped into it. The terminal velocity of the
falling sphere can be measured, and this in tum relates to the fluid's viscosity.
.'I Data:
"
Example 2.3 Apply Stokes' law to the falling sphere viscometer and write an
I Oil density, 60 lbm/ft 3
Oil viscosity, 10 lbm/hr ft
Width of sheet, 5 ft
expression for the viscosity of the liquid in the viscometer.
r Velocity of shee!, 1 ~t/sec
L
-------------------------------------------------~ -------------------------------.- .. "._.~-. .
.,.,--,~-------
,___~....,i Oil
2.6 A wire is cooled after a heat treating operation by being pulled through an open-ended,
oil-filled tube which is immersed in a tank. In a region in the tube where end effects are
negligible, obtain an expression for the velocity profile assuming steady state and all physical
properties constant.
Tube inner radius: R r
Wire radius: KR
Wire velocity : U
2.2 Develop expressions for the flow of a fluid between vertical parallel plates. The plates
are separated by a distance of 2c5. Consider fully developed flow and determine
l a) the velocity distribution,
I'
I
:
b) the volume flow rate.
Compare your expressions with Eqs. (2.20) and (2.23).
2.3 Repeat Problem 2.2, but now orient the plates at an angle f3 to the direction of gravity
2.7 Starting with the x-component of the momentum equation (Eq. 2.45), develop the
x-component for the Navier-Stokes' equation (constant p and 17, (Eq. 2.56)).
and obtain expressions for 2.f a) Consider a very large flat plate bounding a liquid that extends toy = .+ oo. Initially,
ft I jl_
the liquid and the plate are at rest; then suddenly the plate is set into motion with
a) the velocity distributions,
11
1i,I It,_: velocity V0 as shown in the figure below. Write (1) the pertinent differential equation
. I b) the volume flow rate . in terms of velocity, for constant properties, that applies from the instant the plate
I 1 ,
Compare your expressions with the results of Problem 2.2 and Eqs. (2.20) and (2.23), and moves, and (2) the appropriate boundary and initial conditions. The solution to these
try to grasp the "madness behind the question." equations will be discussed in, Chapter 9.
2.4 A liquid is flowing through a vertical tube 1 ft long and 0.1 in. 1.D . . The density of the
liquid is 1.26 g/cm 3 and the mass flow rate is 0.005 lb/min.
a) What is the viscosity in cP? lbm/ft-hr? y v, = 0 v,
L
'\ b) Check on the validity of your results.
r2.~A-I_iquid flows upward through a tube, overflows, and then flows downward as a film ~n Vo
th outside. x % ~
Initially Time I Time 2
a) Develop the pertinent momentum balance that applies to the falling film, for steaoy-state
laminar flow, neglecting end effects. b) A liquid flows upward through a long vertical conduit with a square cross section.
b) Develop an expression for the velocity distribution. With the aid of a clearly labeled sketch, write (1) a pertinent differential equation that
describes the flow for constant properties, and (2) the appropriate boundary conditions.
Consider only that portion of the conduit where flow is fully developed and be sure
that your sketch and equations correspond 'to one another.
'
2.9 a) For flow parallel to a flat plate, develop an expression for the momentum boundary
layer thickness assuming that vx = a sin by, where a and bare constants that must be
evaluated. Apply the approximate integral technique.
b) Apply the approximate integral technique for flow parallel to a flat plate and develop
an expression for the momentum boundary layer thickness, assuming vx = a + by,
where a and bare constants that must be evaluated. Also, write an expression for the
L
drag force exerted by the fluid on the plate's surfa;e for this velocity distribution.
f . ;::- ' .
\ 2.10\ Liquid steel (0. 75 %C) at 2900F is deoxidized by the addition of aluminum which forms
alumlrn (Al ,O,) 'w, c~ obi.In a betto- quality of "'" , '"" alumlrn partldo> are allowOO to
--------------------~---------------------------------~
float to the surface. Determine the size of the smallest particles that reach the surface two
minutes after the steel is deoxidized if the melt is 5 ft deep.
Data: p (steel) = 0.30 lb/in 3
p (Al 2 0 3) = 0.12lb/in 3
3
2.11 Molten aluminum (1300F) is degassed by gently bubbling 75 ;,; N 2 -25 ~;,; Cl 2 gas through
it. The gas passes through a graphite tube (/6 -in I.D. x 3 ft long) at a desired rate of240 in 3 /min TURBULENT FLOW AND
(I atm, l 300F). Calculate the pressure that should be. maintained at the tube entrance if the EXPERIMENTAL RESULTS
pressure over the bath is 1 atm.
Data: p (Al)= 160 lbm/ft 3
= 0.308 ft/sec.
Velocities as small as these (in 1-iIJ. pipes) are not often encountered in practical
engineering, so that most problems of engineering importance occur in the region
of turbulent flow.
In this chapter we shall examine some semiempirical information available
for various systems. We may classify the flow systems into tw,.9.,&i;,qups: flow through
closed conduits and flow past submerged bodies. In flie101iiie't group, pressure-
-drop data are usually desired and correlations given in terms of a friction factor;
in the latter group, forces acting on submerged bodies are usually sought, and
correlations reported in terms of a </:;,~cient.- bv;, c ct d. 5
L
rn-i"'llTDDienrnow-ano-expenmenra1-.-esun~---- - - -~--------------------
Equation (3.3) f6~uses our attention on what should be considered if we ask the
(3.3) I-+ Laminar boundary layer
following question: What pressure drop is necessary to fe1fv4e"'?a" given volume of Distance from wall
fluid through a tube? Thus we learn from Eq. (3.3) that FK must be determined; Fig. 3.1 Velocity distribution for turbulent flow in tubes in the region near the wall.
for laminar flow, this can be calculated because the velocity distribution is amenable~c'.11
to analysis, and pressure drops can be determined a priori. For turbulent flow, 3.1.1 Dimensional analysis for friction factor
the uncertainties involved in parallel analysis have led engineers to take an
experimental approach to the problem. "JV'~~" Now we resort to dimensional analysis which is a method of deducing logical
In turbulent flow we may think that the flow p.Wl~'r; starts with a laminar groupings of the variables involved in a process. One of these methods, called the
boundary layer, in which the flow can be described by Newton's law of viscosity, similarity technique, is applied to systems which are geometrically similar. For
followed by a transition region in which the degree of turbulence stfAAi.YeJr,creases example, in two circular tubes, all comparable lengths have identical ratios. Thus
and laminar effects diminish, until finally the region of fully devetoped turbulence we write
is ~'aC'llect.'- These regions are illustrated in Fig. 3.1. In turbulent flow, therefore,
the fluid still ~~~ 1 jo the solid wall. Thus Eq. (3.1) is applicable, but, in general, (3.7)
the value of r 0 is no( determined by analytical means. More often we employ an
empirical technique and express the force FK as the product of a characteristic area where z 1 and z 2 are comparable distances along the lengths of tube; r 1 and r2 are
A, a characteristic kinetic energy K (per unit volume), and a dimensionless quantity comparable radii in t-ubes 1 and 2, respectively.
f, known as the friction factor: For the flow system in either tube, the differential equation for the conservation
of momentum that applies in the steady state is
(3.4)
This is a useful definition because f is a function only of the dimensionless Reynolds '1[~r ~(r dvz)]
dr dr
dP
--=0
dz '
(3.8)
number fo.r a given geometrical shape. This is shown in a following discussion, but
first let us present how f is related to r 0 . o-O<>'
and a boundary condition is
Fo& flow in conduits, A is taken to be the wft'1urface 2nRL, and K is taken
to be the kinetic energy based on the average velocity, that is, iP V 2 Thus for !rzl r=R
=!o = -11(dV:z)
dr r=R
(3.9)
flow in a filled pipe of length L, we have
ffe11 0 The differential equation and the boundary condition require certain relationships
FK = f(2nRL)(!-p i72). (3.5)
_.,--------,------------------------------- --------------------------~
between velociti~s, dP(d~, and fluid properties, at corresponding points in the two If we define a new ratio as
syst~ms. Thus, m_ ad_dit_10n_ to the ratios of Eq. (3. 7), we define the ratios for kine-
matic and dynamic s1m!lanty:
then the boundary condition, (Eq. 3.9), applied to both systems, yields
or
KP= P1.
(3.10)
P2 Dividing by the Reynolds number, we obtain
To investigate the conditions which must be satisfied by these ratios we t '01 !02
Eq. (3.8) specifically for system 1 : ' wn e ~-=~, (3.16)
P1V1 P2V2
d ( rl -dvzl)] - -dP1 = 0 which are equivalent to half the friction factor (Eq. 3.6). Now the dimensional
171 [ -1 -d (3.11)
r1 r1 dr 1 dz 1 analysis is complete; it has been shown that for flow in the two arbitrary systems
having geometric similarity, the friction factors are equal when the Reynolds
Now if_ we replace r1' Vzb etc., by their equivalents in Eqs. (3.7) and (3.10), then we
can wnte Eq. (3.11) for system 2: numbers are equal. Thus for flow in tubes, the friction factor f may be correlated
as a function of the Reynolds number.
2 - 112 [ - 1 - d ( r1 _z_
KqKv) dv 2 ) ] - Kp -dP2 =O
( - -Kv r 2 dr 2 dr 2 dz 2
(3.12) 3.1.2 Experimental results for friction factor
Multiplying Eq. (3.12) by Kv/K;KP, we finally obtain: Experimentally, f can be measured by noting that if FK is eliminated between
Eqs. (3.3) and (3.5), then "'
and the connection with Eq. (3.15) is made apparent. Also, as we have shown,
It is apparent that Eq. (3.13) written with subscripts 1 would also be valid and
the friction factor is only a function of the Reynolds number: f = f(Rev). In most
because both systems must obey the original differential equation, then '
designs the pressure drop is not u_sually known a priori; we therefore use Rev to
evaluate the friction factor and then the pressure drop developed due to friction.
and
Since f is only a function of Rev, it is sufficient to plof'only a single curve off
or against DV/v rather than determine how f varies for separate values of D, V, p,
and 17. The lower curve in Fig. 3.2 gives a plot of/versus Rev for smooth tubes.
This curve reflects the laminar and turbulent behavior of fluids in long, smooth,
(3.14) circular tubes. For the laminar region,
and 16
f =-3' (3.18)
Re
D1(dPi/dzi) D 2 (dP2 /dz 2 ) which can be derived by using the Hageri:Poiseuille law developed in Chapter 2.
P1 Vf P2 V~ (3.15) The turbulent region has been established solely by experimental data. The
The group_in_Eq. (3.14) is the Reynolds number which, of course, is dimension- entire turbulent curve closely approximates
less. On~:, ag~m ~t appea~s-in_fact, in all cases of forced convection, the Reynolds 1/.jj = 4.0 log(Re.jj) -;; 0.40, (3.19)
n~mbe~ is a s1gmficant d1mens10nless group. The group in Eq. (3.15) is anoth
I
I d1me~s1?nless group and, as shown in the fo:llowing paragraph is actually twi~~ and a simpler expression exists for 2.1 x 103 < Re < 10 5 , namely,
the fnctlon factor. '
f = 0.0791 Re- 114. (3.20)
~--.- ...-,.,...,. ..,,,,.... ..- .. n:Jl'n-R...u-~pit:'rlllll(;II'l.""A"l-:l"~Ull."'----------------~
0.006 ._.,,\ Example 3.1 Determine the mass flow rate (lbm/hr) of water at 80F through
11\ 600 ft of horizontal pipe, having I.D. of 5 in., under a pressure drop of 5 lb/in 2
0.005 ~\ 0.001
The relative roughness s/D of the pipe is estimated as 0.001.
0.004
~' 0.0004
Solution. The solution to this problem can be found by using Eq. (3.17) and solving
0.003 0.0001 for the average velocity V. However, V cannot be determined unless f is known
and, because! = f(Rev), a trial and error solution is in order.
0.00005 5
First approximation. In Fig. 3.2, f = 0.005 when s/D = 0.001 and Re = 9 x 10 .
0.002
JO' ]04 10 5
10 6
10 7 10 8 For this friction factor, the velocity is
Re= VD/v
Fig. 3.2 Friction factors for flow in tubes. (Adapted from L. F. Moody, Trans. ASME66, 671
(1944), and Mech. Eng. 69, 1005 (1947).)
Table 3.1 Values of absolute roughness e. for new pipes
5
D ft 12 1
Ft In. I= 600 ft= 1440'
Drawn tubing, brass, lead, glass, centrifugally
spun cement, bituminous lining, transite .. . 0.000005 0.00006
l'lP = 5 lq/in 2 = (5)(4633) lbm ,
ft sec 2
..r ' ,
Commercial steel or wrought iron .............. . 0.00015 0.0018
Welded-steel pipe................................... . 0.00015 0.0018
11P (5)(4633) lbm ft 3 3600 2 sec 2
4 81 x 10 9 ft 2 /hr 2 .
p ft sec 2
62.4 lbm 1 hr 2 =
Asphalt-dipped cast iron .......................... . 0.0004 0.0048
Galvanized iron ...................................... . 0.0005 0.006 Then
Cast iron, average................................... .
Wood stave............................................. {
0.00085
0.0006 to
0.0102
0.0072 to
--j(l)(l)
V- o.oi
1440
(4.81 x 10 9 ) = 1.83 x 104 ft/hr.
0.003 0.036
0.001 to
Concrete ....... ~ ..................................... { 0.012 to This velocity gives the following Reynolds number:
O.Ql 0.12
4
0.003 to 0.036 to Rev = DVp = 5 ft 1.83 x 10 ft 62.4 lbm hr ft
.Riveted steel. ..... .,................................... {
0.03 0.36 1] 12 hr ft 3 . 2.25 lbm
e e m ft e in in:
Note: - = -- = ----
D D in ft diam. in in. = 2.11 x 10 5 .
---------------------..---~=--------------------------.-."TT""""l'u.., ..--;:yu.,:o:.,.-6"-.._-....,,_ ._.-o-~v
Second approximation. For Rev= 2.11 x 10 5 , f = 0.0052. Thus, with this f, V and is given by Fig. 3.3. The laminar-to-turbulent transition for noncircular
is slightly corrected: conduits is still at a Reynolds number of approximately 2100.
112
-
V (0.0050) 4 zi/z 2
= 0.0052 (1.83 x 10 ) = 1.82 x 104 ft/hr. "'
0:: 1.2 0.5 0.6 0.7 0.8 0.9 1.0
W= 104 )
Here we evaluate the Reynolds number using De, which for rectangular ducts is
2(z 1 z 2 )
D = '
e (z1+z2)
* Many other texts refer to the hydraulic radius Rh rather than equivalent diameter. In such v_
instances, Dis replaced by 4Rh, where Rh = A/Pw; the end result is the same whether De or Rh Fig. 3.4 Laminar and turbulent boundary-layers next to a flat plate (the vertical scale is
is employed.
greatly exaggerated).
___l
-..--.--------,.~rurnmenrnow-anu-expenmenra1-resu11s-- ---------------~~,------..
We again see that the friction factor is a function only of a Reynolds number.* These ~quations are valid for a fl.at plate at zero incidence to the flow, providing
These results hold so long as the boundary layer itself remains laminar. However, separation does not occur.
at a value of Rex between 300,000 and 500,000 the layer becomes turbulent,
Example 3.2 In order to study the mixing action in a molten bath of aluminum
increasing significantly in thickness, and displaying a marked change in velocity
contained in an induction furnace, we use a steel flag. The flag is held in a vertical
distribution. We depict this transition in Fig. 3.4 which shows a much steeper
position, and placed in the central part of the furnace where the metal flows
gradient near the wall and flatter gradient throughout the remainder of the boundary
upward, as depicted in the figure below.
layer for the turbulent zone. As a result, the shear stress at the wall is greater in the
turbulent layer than in the laminar layer.
Using the applicable differential equation, the geometric similarity argument
would again indicate that the friction factor is a function of the same dimensionless
Induction
group, Re. Again, experimental data have been gathered to correlate the friction
factor and the Reynolds number. The results in Fig. 3.5 show that the laminar ~ls 0
region may extend up to a Reynolds number of about 106 . However, this condition 0
is only achieved on smooth surfaces, and if the system is protected against external 0 0
disturbances, such as minute vibrations caused by marching protestors, passing 0 0
trucks, or bomb explosions on other parts of the campus. The results can be 0 0
represented as a simple function for a limited range: 0 0
0 0
115
7 0 0
f = 0.072(-v-) ,
5
5 x 10 < Re < 10 (3.25)
0 0
VOOL
0 0
A somewhat more complicated function 1 exists for the entire graph: 0 0
f = 0.455 (3.26)
Arrows indicate the melt's motion
(log Re)z.5s due to the electromagnetic forces.
Boundary layer If provisions are made to measure the added force exerted on the flag due to
Entirely Transitional Entirely
fluid motion, prepare a control chart that relates this force to the velocity of the
f laminar turbulent molten aluminum in the central portion of the furnace. Assume that turbulent
0.010
0.009 .flow encompasses the conditions of interest.
0.008 -rzzzzzzzz22z4
0.007 Flow~L--..1
0.006 Data: Viscosity of aluminum, '1 c~,
0.005 Density of aluminum, 160Ibm/ft 3 ,
0.004
Flag dimensions, 1 ft x 1 ft x 0.01 ft.
0.003
Solution. In this problem it is convenient to choose values of ReL and then cal-
0.002
culate the corresponding sets of FK and V00 The velocity of the aluminum is
found by:
0.001.,,_,...__,__Lc-'--f~...._,_,~~,,...._~~~'--L~----'--'
Y/
10 4 105 106 107 108 10 9 10 10 V00 = -ReL,
ReL
Lp
Fig. 3.5 Friction factor for flow parallel to fiat plates. (From J. K. Yennard, Elementary Fluid and then we can write the corresponding force exerted by the moving aluminum
Mechanics, Wiley, New York, 1954, page 384.) ori both sides of the flag:
'
* Most texts on fluid mechanics use the term drag coeffi.Cient CD rather than friction factor f, FK ~ 2f(L W)(tp V~)
when considering flow past submerged bodies. When the form drag is entirely friction drag,
an even more specific term, skin friction coefficient C1 is sometimes used.
1
H. Schlichting, Boundary Layer Theory, McGraw-Hill, New York, 1955, page 439. = fLWp(L;: 2 Rei)
= f( ~)(~)(~ 2 Rei). We have found that this force, often termed the form drag, is composed of two
parts: the friction drag and the pressure drag. In a well-streamlined body (flow
For ReL = 10 6 , Fi,.-3.5 gives f = 0.0045. parallel to a flat plate, airplane wing, hull of a boat, etc.), the friction drag not only
forms the major part of the total drag, but may even comprise it entirely. In the
1 Voo ~ 2.42 lbm ft
3
1 hr 106
= 4.20 ft/sec, case of a body with sharp corners, or a sphere, or a cylinder oriented normally to
hr ft 1 ft 160 lbm 3600 sec flow, separation occurs, and a turbulent wake forms.
and With a plate oriented perpendicularly to flow, the separation always occurs
at the same point, and the wake extends across the full projected width of the body,
F. 0.0045 1 ft ft 3 2.42 2 Ib;, 1 hr 2 sec 2 Jb1 110 12 as shown in Fig. 3.6; this results in almost all pressure drag comprising the form
2 2 2 2
K - 1 ft 160 lbm hr ft 3600 sec 32.2 ft lbm drag. If the body has curved sides, the location of the separation point is deter-
,~)
= 0.382 lbf. mined according to whether the leading boundary layer is laminar or turbulent,
as depicted in Fig. 3.7. In turn, this location determines the size of the wake and
In a similar manner, the corresponding values of V00 and FK can be found for other the amount of the pressure drag. After this brief introduction, let us now consider
values of ReL: some of the available data.
i V ft/sec
00 , -FK, lb1
I
105
l
0.420 0.006
5 x 10 5 2.10 0.108
106 4.20 0.382
2 x 10 6 8.40 1.36
4 x 10 6 16.8 4.89
5 x 10 6 21.0 7.22
I
(b)
20 (a)
Velocity of melt, ft/sec Fig. 3.7 Separation of boundary layer and wake formation with a cylinder or sphere immersed
in a flowing fluid when (a) the leading boundary layer is laminar, and (b) the leading boundary
~ 3.3.2 Flow past a sphere and other submerged objects layer is turbulent.
In Chapt~r 2, we analyzed laminar flow past a sphere, and gave the kinetic force
of the flmd exerted on the sphere: For the presentation of data, we use Eq. (3.4r As we come across the friction
factor again, let us repeat the expression
(2.123) ,FK = fAK; (3.4)
-~------------------------~-r--~-----~~-~-----
K still remains tP V~, but for shapes such as spheres, cylinders, ellipsoids, etc., we
choose the area A as the projected area normal to the flow. For example, for flow Plate l
Infinite square to flow
past a sphere, FK takes the form cylinders
L Infinite
(3.27) v_ - L~ v_ - r-~\ circular
[_ ~ _ -=- _ cylinder
From Stokes' law, which is the result of an analytical solution for laminar flow v_._v~___:::::i.-----
r v__ L~ D~
around a sphere, we can write the following equation by rearranging Eq. (2.123):
1
8
Infinite ---------~--
- -----------, "J7
circular
FK = ( nR 2 )(zpV
1 2 ( 24 )
00 ) - - (3.28)
6
cylinder Finite---=-:_:;r- ...- D '.1,
DV00 /v circular l__=-:::r-..... j \.
cylinder v_- ~ .
By comparing Eqs. (3.27) and (3.28), we obtain the friction factor (drag coefficient) 10-
LjD=5 t
I :4 Elliptical
for laminar flow past a sphere: cylinder
f = 24. (3.29)
Re
2
10- L__,_
_ 2 _ _,__~-+--+--'-~!---+--''--=-*---!~""'-,;--+---!--'--+--+---'-~i---7-:~
s s s 10 3 2 s 2 s s 10
Once again, the,friction factor is a function of the Reynolds number. This is also 10 1 10
2
10 ~ 2 10 10s 2
true for turbulent flow but, as you may suspect, Eq. (3.29) does i;iot apply for all Ren =DV_/v
velocities, and experimental correlation is sought in the form of f = f(Re). Fig. 3.9 Friction factors for submerged bodies. (Adapted from F. Eisner, Proc. 3rd Intern.
Figure 3.8 shows such a correlation for spheres, as well as for some other geometrical Congr. Appl. Mech., 1930, page 32.)
shapes. We also include Fig. 3.9 to indicate other possibilities. Note that Stokes'
law applies up to Re ~ 1 for spheres.
The phenomenon of boundary-layer separation has a major effect on the form
drag. This effect is noticeable at large Re where the form drag comprises most of
the total drag. At this high Re, the transition to turbulent flow occurs before
/102
8
6
separation, and the separation point is moved farther downstream, as illustrated
in Fig. 3.7, thereby reducing the form drag.
2
In metallurgical processes, the discussed principles of flow past submerged
bodies mostly apply to situations in which the body moves in a fluid. The friction
10 -~.
factors are.determined experimentally, and it makes no difference whether the
"~. Ellipsoid I : 0. 75 body moves in a still fluid or is held stationary while the fluid flows past it. How-
'~
Stokes' law~
.... _ L DDisk
v_ -v
v_ ...n(J
ever, when the body is irregularly shaped, friction factors which are measured
- - - - - _ y_:_ \... _ - - - - - Ii--- while the body is stationary may not be applicable when the body moves, because
during free motion the orientation of the body may change rapidly and repeatedly.
-1::. -- -.
~-TDC)
. ~ It is often desirable to separate particulate mixtures of solids. These mixtures
Sphere D ' may consist of particles of the same material which have different sizes (homo-
';
v v--c;;;,"
- T- ~.
'
\
Ellipsoid 1 : 1.8 -
.- geneous mixtures), or they may consist of particles of different materials and various
sizes (heterogeneous mixtures). Homogeneous mixtures can be separated by
screening, but for separating mixtures of fine sizes, free settling or centrifugal
methods are preferred. In general, screening does not work for heterogeneous
mixtures, and therefore separation is gained by free settling or centrifugal methods.
Ren =DV_/v
Example 3.3 In powder metallurgy operations, it is important to separate homo-
Fig. 3.8 Friction factors for submerged bodies. (Adapted from F. Eisner, Proc. 3rd Intern. geneous mixtures in the range ofparticle sizes from 1 to 50 . For this purpose we
Congr. Appl. Mech., 1930, page 32.)
apply the principle of air elutriation, which is illustrated by the devic.e on page 90.
--------------------
Elutriated particles to the This is a very high velocity; thus, for all situations, Rev < 1, so that Stokes' law
fines collector
does apply. According to Eq. (2.125),
D-
-
vi l8V<Xl17 .
(p. - p)g
Here we may ignore the fluid density p.
For V <Xl = 10 cm/sec,
,
" (18)(10)(1.80 x 10-4) = 0 00287 =
D= (4.0)(980) . cm 28.7 .
Suspended particles
We can calculate values of D for other air velocities in a similar manner; the results
are presented in the graph below.
Air
in
The force exerted by the air stream is great enough to suspend particles of a given
diameter, and all particles of smaller diameter are carried upward and to the
collector of fines. Larger particles fall back against the air stream and down into a
settling chamber. For the effective operation of this device it is necessary to know
the size of particles suspended at a given flow rate.
Assuming that a spherical particle, with a specific gravity of 4:0, applies to a ('
40 50
homogeneous mixture of iron powder, draw a graph that relates the diameter of
suspended particles to the velocity of the air in the expanded portion of the tube. vro, cm/sec
Solution. To obtain the solution, recognize that particles which have terminal 3.4 FLOW THROUGH PACKED BEDS OF SOLIDS
velocities equal to the upward air velocities are those which remain suspended.
Further, examine the criterion that Stokes' law applies for Rev < 1 and that a Packed beds of granular solids or agglomerates of fine particles occur in many
simple expression for the diameter may be appropriate. metailurgical processing systems, ranging from sinter plants for agglomerating
iron ores to the production of intricate parts via powder metallurgy. The flow of
For air at 80F, the fluid properties are 17 = 1.80 x 10- 4 g/cm sec and p = fluids through. porous aggregates is not simple, especially due to the effect of the
0.00118 g/cm 3. For Rev= 1 and D = 5 x 10- 3 cm (the largest particle size of properties of the aggregate on the flow. But under certain conditions we may
interest), the corresponding velocity is predict the flow reasonably well.
where Q = volume offtuid flowing per unit time, cm 3 /sec, A = cross-sectional area, Engineers usually know V 0 rather than V; V0 is called the superficial velocity
cm 2 , and kn = permeability coefficient, a constant depending on the fluid, tempera- and is defined by
ture, and packing characteristics, cm 4 /dyn-sec. Equation (3.30), which is known
as D'Arcy's law, has been applied to the problem of the permeation of gases V0 = Q/A, (3.33)
through foundry molding sands, and the filtration of water through filter cake. and since
The kn in Eq. (3.30) is often called the permeability, which is satisfactory as long
as we carry out the test with the same fluid at the same temperature, but, in general, V0 = flw, (3.34)
it is more common and more desirable to define a specific permeability f1J> by then
means of
(3.35)
t~1.:JrV~ 1 ;""'1, (3.31)
/
1: ' where w = void fraction.* The concept of the hydraulic radius Rh was introduced
where Y/ is the viscosity of the fluid. This allows f1J> to be specific to the packing only, with the equivalent diameter (Eq. 3.21).
and therefore allows the flow of other fluids, or the same fluid at other temperatures, For packed beds, Rh is defined as
to be predicted. The units of specific permeability f1J> are
average cross section available for flow Ah
(d;::ec)(d~:~ec)
Rh= = - (3.36)
q> = = cmz, average wetted perimeter Pw
By the average wetted perimeter, we mean the average total boundary line between
where the unit of specific permeability, the D'Arcy, is now defined as the fluid and the packing, viewed by a slice through the bed normal to the axis of
1 D'Arcy= 1 x 10-s cm 2 the bed. Thus, in a bed of length L,
Unfortunately, this value has not been standardized and various other definitions volume available for flow AhL
Rh = ---------- (3.37)
of the unit have- been proposed. The British Standard Specifications define the total wetted surface PwL
D'Arcy as expressed with viscosity in centipoises and the pressure drop in atmos-
Here AhL is also the volume of voids in the packing, and if both the numerator
pheres. This D'Arcy differs from the above definition by 1.3 %, since and denominator of Eq. (3.37) are normalized by the bed volume, we obtain
1 D'Arcy (BSS) = 0.987 x 10-s cm 2 .
AhL/V w
Rh= PwL/V = s' (3.38)
~ 3.4.2 Tube-bundle theory and Ergun's equation where w = void fraction or voidage and S = S0 (1 - w).
The factor S 0 is the total surface area of particles per unit volume of particles,
D'Arcy's law is an empirically observed law. However, a semitheoretical approach while S is the total surface area per unit volume of bed. From this,
to the problem yields D'Arcy's law with more insight into the effect of the packing
on f1J>. The theory, which is known as the tube-bundle theory, regards the bed as a w
Rh= . (3.39)
bundle of tangled tubes with weird cross-sections. It is assumed that the packing S0 (1 - w)
is uniform without isolated porosity, that there is no channeling of flow, and that
We may substitute Eq. (3.39) into Eq. (3.35) and obtain
the diameter of the column is much greater than that of the particles. To start with,
we look to the Hagen-Poiseuille formula for laminar flow in Chapter 2 (Eq. 2.33). f).p w3
By analogy, we say that Vo= Ki LriS5 (1 - w)2. (3.40)
_ !).PR~
, :It V= K i - - ' (3.32) Equation (3.40) which is the form of the pres~ure-drop relationship for flow
'I
r
Lri
*We use the symbol where to denote the void fraction, instead of the more common e or P,
where Ki = constant of proportionality, and f7 = average velocity in the inter- order to avoid confusion with the emissivity e, which is introduced in later chapters, and
ill
stices of the bed. With the pressure P used in this chapter.
through packed beds, is valid for the lower range of Reynolds number (laminar 8 JOO
flow region), where K 1 1 has been found to equal 4.2. 2 Insertion of this value into I
M8 :::'.- 50
Eq. (3.40) gives ~~
;:,.
1liP w3 .3"
J
0 i,=--- ' (3.41) II
4.2 L71S6 (1 - w)2 ~ IO
which is the Blake-Kozeny equation. Note that this is the same as t!v.e D'Arcy i 5.0
equation, where
(3.42) 1.0
or 0.5
OJ3
f!J> = . (3.43)
4.2S6( 1 - w )2 0.1 L__ _ _ ____,_,__ _ _ _~_ _ _ _ _,.,,..,,._ _ _ _-,-'
0.1 1.0 IO JOO !000
These equations emphasize the fact that kn depends on the properties of both the
fluid and solid, while f!J> depends only on the properties of the solid phase. Re - pVo
' - l'f(I - w)S 0
According to the flow behavior in all forced convection systems, a fluid velocity
is eventually reached beyond which laminar flow no longer prevails. Under these Fig. 3.10 Pressure-drop correlation for flow through packed beds. (Based on S. Ergun, Chem.
conditions we again resort to the use of a friction factor which can be correlated Eng. Frog. 48, 93 (1952).)
solely as a function ofa Reynolds number. For packed beds, the modified friction
factor may be measured by
When applying this expression to gases, we take the density of the gas as the
!c = LS 0 pV6(I - w). (3.44) arithmetic average of the end densities. For large pressure drops, however, it is
probably best to apply Eq. (3.46), to express pressure gradient in the differential
This is analogous to Eq. (3.17). We utilize a modified Reynolds number for packed form, and to integrate over the bed thickness, taking into account the variations
beds in the correlation in density, viscosity, and superficial velocity.
Thus far, we have. defined the modified Reynolds number and friction factor
Re = pVo (3.45) in terms of S0 ; the particle surface area per unit particle volume. If the particles
c 71(1 - w)S 0
in the bed are sphere of uniform size, S0 may be easily related to the diameter, Dp,
When the flow exceeds that for Rec ~ 2, then Eq. (3.41) no longer applies and by
we use Fig. 3.10. The equation
(3.48)
fiP 4.271V0 SW - w) 2 0.292pV02 S0 (1 - w)
y = w3 + w3 (3.46)
Then Eq. (3.46) becomes
describes the entire curve analytically; the second term on the right describes the
pressure drop under highly turbulent flow conditions. In dimensionless groups,
,-r;p
~=
15071V0 (1 - w) 2
+
1.75pV02 (1 - w)
'
\
(3.49)
L Df,w 3 . Dpw 3 ,'
.-"
4.2
!c = - + 0.292. (3.47) which is known as Ergun's equation. It is then useful to define ;-ne~ friction factor
Rec which is often associated with Ergun's work:
2
S. Ergun, Chem. Eng. Pfog. 48, 93 (1952). The constant 4.2 is not universally selected; some
believe the value to be as high as 5.0, based on a paper by P. C. Carman, Trans. Inst. of Chem.
Dp fiPw 3
Eng. 15, 150-166 (1937). fE = 6fc = LJ'Vo2 (l _ w)' (3.50)
L
_ __,-------"""-.-a:~~11.-~~a1n.~~1-1I11ic:11t:a1-1~u11~-------------------
and a new Reynolds number the arithmetic mean of the openings of the two screens defining the corresponding
rnass fraction, A</>;, collected between them, then
DppV0
ReE = 6 Rec = ry(l - w)
. (3.51) 6 n A<j>i
Sw =/'~L '=-' (3.53)
PP}= 1 Dp,
Now, in dimensionless form, Eq.(3.49) becomes
where n is the number of screens used, and pp is the density of the particulate
150
fE = -R + 1.75. (3.52) material. The average particle diameter for this m"ixture is called the volume-surface
eE mean diameter, defined as
When a bed contains a mixture of different-size particles, the question arises - 6 (3.54)
D =--
as to what should be used for Dp in Eqs. (3.49)-(3.51). To this day, the question is vs SwPP
/
still unanswered, but a suggestion commonly put forward (although not adequately //
tested) is to use the volume-surface mean diameter, Dvs. This parameter is discussed Noting that S0 = Swpp, then /\'
in the following paragraph. 6
For a bed containing a mixture of different-size spherical particles, S 0 may be So =T\ (3.55)
determined from the specific surface of the mixture, Sw (cm 2 /mass of particles). '"1!.Y
If the material is screened, and the diameter of each size fraction, l5 P,, is taken as and, by comparing Eq. (3.55) with Eq. (3.4.8), the proper value of Dp to use is Dvs
Finally, the situation arises that we have a bed of nonspherical particles. In
Table 3.2 Shape factors for screened particles this case, we define a. shape factor A., which does not depend on particle size and
which is a function of the shape of the particles only, by:
Substance Type of measurement Shape factor, ii. /.-~- ''. /''
Calculate the pressure drop, prior to ignition, across a bed of sinter 12 in. deep 0.6
(w = 0.39) for air flowing at 60F, and with V0 = 25 cm/sec. The surface area S
0.4 '----'-------"'---'---'--~-~-~~.....,.......,.,__,
measures 8!_cm 2 /cm 3 . 1 2 4 6 8 10 20 40 60 80 100
Tube-particle diameter ratio, DT/Dp
Pair = 1.23 X 10- 3 g/cm 3 (STP), Fig. 3.11 Correction factor for wall effect. (Derived by H. E. Rose and A. M. Rizk, Proc. Inst.
1'/air = 178 x 10- 6 poise (60F). Mech. Engrs., London, 60, 493 (1950).)
I
c
we use D'Arcy's law (Eq. 3.30) in its simplest form, and the permeability kv is
Using Eq. (3.46), we get obtained from
2 I
I
2
AP 4.2
pV0 S( ) 3007.2( cm )
(3.58)
L = ~ Rec + 0.292 = 2
2385 (dyn/cm )/cm kv = - t - min-cm H 2 0 .
= 2.04 in. H 2 0/in. bed; The units are common only to this test. Sands with high permeability to gases
havt' kv values in the range 300-500; a typical value would be 60-120 cm 2 /min-cm
AP = 2.04 x 12 = 24.45 in. H 2 0.
H 2 0.
The measured value in the case studied 3 was 25 in. Voice, Brooks, and Gledhill 4 studied the relationship between pressure drop
across a sinter bed and the rate of flow of air and found that
3.4.3 Wall effect
It is probably obvious to the reader that the container diameter must be a good deal Vo=~= kv(A:)-", (3.59)
larger than the mean particle diameter of the packing, in order for the above equa-
tions to be valid, since the void fraction at the wall will be larger than the bulk value where kv is the permeability constant, and n is a variable which depends on the
of w. This wall effect is demonstrated in Fig. 3.11 which shows that the container stage of the sintering process, but which averages 0.60 for the process as a whole,
diameter should be approximately 20 times the particle diameter in order that as long as extra-fine ore is not used. For the cooling of hot sinter, values of 0.53
Eq. (3.46) could predict the pressure drop to within 10% for Rec between 2 and 150. have been found for n. If one considers Eq. (3.46) and assumes turbulent flow,
If necessary, we may first compute the pressure drop using Eq. (3.46), and then use that is, only the second term on the right-hand side is significant, then one finds
Fig. 3.11 for correction.
_ (
Vo -
w3 )o.5(AP)o.5 -_ kv. (AP)o.5 ,
3.4.4 Applications 0.292pS 0 (l - w) L L
In practice, several tests have been developed for control of the permeability of
I which is in good agreement with the observed values, considering the assumptions
aggregates. In foundries, a standard test subjects a cylindrical sample of sand 2 in.
3
I made.
4 E.W. Voice, S. H. Brooks, and P. K: Gledhill, J.S.J. Spec. Report 53 (1955).
l_
D. W. Mitchell, J. Iron & Steel Inst. 198, 358 (1961).
. In powder metallurgy, the permeability of a sintered metal compact is of in shape, and exhibit a relatively high degree of friction between particles. Typical
importance because it is a measure of the relationship between processing variables values of w lie between 0.35 and 0.5, except, of course, where pressure is applied
and the pore st.ructure developed during sintering, and because, in some instances, in order to force the particles together. In most instances, we must measure w
the fl.ow of a flmd through a porous sintered metal part is important. Self-lubricating in situ, using Eq. (3.62).
beanngs and porous metal filters for. air, water, etc. can serve as examples. Since there are voids in the packing of equal-size spheres, small particles may
5
. Morgan has proposed an equat10n for steady flow conditions through porous enter without changing the overall volume of the bed, so clearly, the size distribu-
smtered metal: tion of the particles is going to have an effect on the bulk density of the bed. Furnas 6
f).p Y/ V0 pV2 has made some classical studies on the void fractions in packed beds, using binary
-=-+ - -0, (3.60) mixtures of particles (two different sizes) in various proportions. He started, in
L </J <!>
each cas~, with materials with the same initial (single component) voidages w 0 ,
\ ~! where </J is called the viscous permeability ' and <!> the inertial permeability:' v;Q, n.,, mixed them, and. measured w for the mixture. Figure 3.12 shows his results for
an~ P ~re defined as before. We should compare this equation with Eq. (3.46), from w0 = 0.5. It is clear that, as the difference in the particle size increases, lower and
which 1t appears that lower w values are obtainable, with minimum voidages occurring in the range
_,
w3 w3 "" 55-67 wt% of the larger-size material. Essentially, the same range for minimum
</J = 4.2S6{1 - w )2 =f?IJ and <!> = - - - - - -
0.292So(l - w)
voidages is found when w 0 = 0.6 and 0.4.
Furnas has also made calculations of the minimum void fractions for three- and
No~, at low values of w, S0 is the surface area seen by the fluid (since there four-component mixtures of particles in which each component alone exhibits the
may be mternal porosity which is not connected to any flow channels and which same voidage. If the coarse and fine particles in a binary mixture are of equal
contributes to thew determined in situ, but whose w surface area doe; not contri- particle density and equal initial voidage, then when the voids, w, in the coarse
bute to the S0 which affects the flow). Because w is then too large S measured in material are saturated with fines, the volume fraction of coarse material is 1/(1 + w),
this manner will be larger than the true value of S0 . Thew used ~h;n the isolated and the amount of fine material is 1 - [1/(1 + w)]. A third, still smaller, component
pore volume becomes significant should only be the interconnected void volume can then be added to the binary mixture, filling the interstices of the second com-
fraction. ponent, etc. The total volume fraction of solids in the mixture is then
(. 1. . [ <- "-V'{.<,"/rft/.J.t?e.;.,'Ji.,I
3.4.5 The void fraction ~pelt"'-.. ttitl~ ;fa cc )Jt~l, atf
An important parameter in (fhe above equations for flow through packed b:c;s is (1 - wm) = _1- + ( 1 - _1-) + ( 1 - _1-) ( 1
~+w 1+ w 1 + OJ (-1-)
-th) + ... ,
the voidage w, which is often difffcult to know or predict under industrial conditions.
We define the voidage, or void fraction, by 1 +OJ
j
-.--------------- ------------- - ---- - - - - ------------------------~
Table 3.3 Effect of size gradations on properties of a rhombohedral packing. (From White
and Walton, ibid.)
Mixture with:
Property Primary Secondary Tertiary Quaternary Quinary
.9 0.6 at a certain point, the bed expan~s slightly, and the individual particles become
ti supported in the fluid stream with freedom to move relative to each other. At this
<!::"' 0.5
-0 point, the bed is no longer static, and is said to be fluidized.
' We relate the pressure drop for the start of fluidization (incipient fluidization)
e: 0.4
" fo the weight of solid particles supported in the fluid stream by
.
.s 0.3
;:;; AP
L = (Ps - p)g(l - w), (3.64)
0.1 where Ps = density of solid particle, and p = density of fluid. Up to the point of
fluidization, however, we give the pressure drop by a relationship such as Eq.
10-s 0.001 (3.57a), and when (3.57a) and (3.64) are equated, we obtain, after rearranging,
0.01 0.1 1.0
Smallest diameter /largest diameter
D~(Ps - p)pg
F(Figr.o3m.1F3 Calcu!ba~ded) minimum voidage in two-, three-, and four-component particle mixtures (3.65)
urnas, 1 1 . 112
By analogy with flow past submerged spheres, we can define a Reynolds number in
~o a reduction in the ove:all void fraction as smaller particles are added. With an fluidized beds: ~-
~d~~l five-component mixture, the voidage would be 0.149, decreased from the
m_itial 0.259. Table 3_.3 i~dicates some of the results of their calculations. Experience I ~= D_--~~_o_P , (3.66)
with_ foundry sands m_d1cates that this approach, although idealized, is useful. For
details of the calculat10ns, the reader should consult the original paper. \ 1502 2 (1 - w) , 1-.75~, , D~(Ps - p)pg
--~3-- Rev + - - 3 - Rev2 = 2 . (3.67)
w w 11
i.., 3.5 FLUIDIZED BEDS !
l Wen and Yu 8 have shown that there ar~ two empirical relationships between the
For an up-?raft packed bed system, an upper limit exists for the fluid flow rate 't shape factor A and the void fraction at minimum fluitlization Wmr:
Let us consider what happens ifth~ flow rate through a static bed steadily increases:.
l
l.. .
I
_,., The pressure drop across the bed mcreases steadily with increasing flow rate until,
I 8
C. Y. Wen and Y. H. Yu, Fluid Particle Technology, Chem Eng. Prag. Symposium Series,
r ,r r
~ 109
W2
Ci:
I M
t f
ff
s... "'" 10
"
8
/':,P
~ t::,p t::,p f::,p w=I
91 ,
1. 1.
wfb
II
"
C) J07 l l
J06
vtb Vw, Vw=1
(Full bed) (Minimum (Particles
fluidization) entrained)
10 5
+
"-
<J
bJ)
104 .Q
10 3
Fig. 3.15 Schematic representation of the relationship between the void fraction in the
10-3 10- 2 10- 1 2
fluidized bed, the superficial velocity, and the pressure drop across the entire reactor, for
10 10 10 3 104 particulate fiuidization.
R' _DpVomrP
eDmr----
11
Fig. 3.14 Generalized correlation for minimum fividization velocity. (Taken from Wen and ' ' In Section 3.3.2 we discussed friction factors for flow past individual particles.
Yu, ibid.) In a fluidized bed, the particles, although suspended by the gas stream, exhibit a
varying friction factor as the voidage varies, until finally at w = 1 the friction factor
-~-------------.,..-...,..~-~T-..----""..,..Y"'-...,..:na-a.,..-.,ua~n
is the same as that for free-fall conditions. Wen and Yu found the relationships of sulfide ores, the reduction of oxides, halides, and sulfides to metals, the drying
between the friction factor and OJ, and expressed their results analytically: of coal and sand, the heat treating of small parts, and the coating of pipe by
fluidizing the organic coating material around the pipe.
47e
OJ .Ga = 18 Rel:i + 2.70 ReJ
687
. (3.69)
Example 3.5 In developing a method for the recovery of uranium from spent
This is shown in Fig. 3.16. Thus, for a given Rel:i value, one may calculate the bed
voidage, or for a given bed expansion and particle size, one may calculate the nuclear fuel elements, engineers have conducted research on the feasibility of
required superficial velocity. Finally, at OJ = 1, the terminal velocity is reached: fluorinating U 3 0 8 to UF 6 in a fluidized bed reactor. 9 The-U 3 0 8 part~les are
mixed with Al 20 3 particles (90 % by volume) and fluidized with N 2 at 750K, to
JI; =J~ Dp(p.~ p)g, (3.70)
which a small amount of F 2 has been added. In order to start an experiment of
3 (!el this type, we must make some estimates.
First, we estimate the minimum fluidizing velocity. Since most of the particles
where f is the friction factor for flow past the particles.
are Al 20 3 , we make appropriate calculations for the Al 20 3 , and then see whether
the velocities are satisfactory for the U 3 0 8 particles.
c3 10
7
The appropriate constants and data are
p(g/cc)
106
Fig. 3.16 Generalized correlation between the particle, bed, and fluid characteristics for a Using V0 mr = 1.17 cm/sec, we obtain
fluidized bed. (Taken from Wen and Yu, ibid.)
R , (6 x 10- 4)(1.17)(4.8 x 10- 4) _3'
evmf = 3.35 x 10-4 = 1.00 x 10
Fluidized beds allow more uniform heating and fluid-solid contact than fixed
beds. They also offer the advantages of continuous operation since they may be which is greater than that for minimum fluidization if Ga = 7.53 x 10- 3 . There-
continuously fed and the product withdrawn ad infinitum. For operation in this fore, 1.17 cm/sec is the minimum fluidizing velocity. -
condition, the superficial velocity has to lie between that given by Eq. (3.68) and
Eq. (3.70).
9
Applications of fluidized beds to metallurgical .processes include the roasting Argonne National Laboratory Report 6763, Oci. 1963.
Next, calculate the velocity for the operating condition when the bed voidage the bottom of the tower is 20 lb1 /in 2 and the bed porosity is 0.40. The gas has a viscosity of
is 0.8: 0.1 Ihm/hr ft at 1000F and atmosphere pressure; the density of the gas at 1000 is approximated
by
w4.7 Ga<At203J = (0.8)4.7(28.2) = 9.9.
p(lbm/ft 3 ) = 0.05 P (atm).
From Fig. 3.16, '!>/
,.; J,\\L t
as flows through at a rate of 753,600 Ihm/hr, what is the inlet pressure?
2 4
Re'<A1io,> = 0 _5 = (1.2 x 10- )(4.8 x 10- )V0 , bed of solids is 60 ft high and 15 ft in diam, with material A forming a central column
D 3.35 X 10- 4 ameter 10 ft and material B filling the annulus between A and the outside. The pressure
V0 = 29.2 cm/sec. at the top of the bed is 10 psig and that at the bottom is 25 psig. For the following character-
istics, calculate the fraction of the gas that passes thr9ugh the A material.
Under these conditions, there is elutriation of the U 3 0 8 , but when the U 3 0 8 A : w = 0.40, Dp = 3 in.
particles are fed into the bed from the bottom, the residence time within the bed B: w = 0.25, Dp = 0.75 in.
is long enough for the reaction to take place, and the U 3 0 8 is converted to UF 6 !-vE.
_y' ,..,...:
before it reaches the top of the bed. Whatever particles reach the top of the bed,
are necessarily blown out, since + ++++++
v 6 -
4 2
- (6 x 10- ) (8.3)(981) - B~
1C U 30s) - (l )( _ x _ ) - 0.485 cm/sec.
8 3 35 10 4
Here, the terminal velocity has been determined by substituting { = 24ri/Dp YiP
into Eq. (3.70). Therefore, if the reactor is operated at 29.2 cm/sec superficial
velocity, the Al 2 0 3 is fluidized, and the U 3 0 8 is entrained when it reaches the top
of the bed. The critical factor which determines whether the method works or
not is the tim_e it takes a U 3 0 8 particle to work its way up through the Al 2 0 3 bed.
+++++++
PROBLEMS
Assume that the temperature of the bed is uniform across the bed at a~y height L, that is, Pga
'.' - -------J
is the same in both columns. Assume that turbulent flow conditions pr~vail.
3.1 Water at 80F flows through a brass 1tube 100 ft long x !-in. I.D. at a rate of 50 gpm.
3.8 Gas flows through a packed bed 10 ft x 10 ft square cross section x 46.3 ft long. An
Calculate the pressure drop (in psi) that 'accompanies this flow.
inlet pressure .of 15.1 psia is measured when 200 lbm/hr of the gas flow through the bed at
3.2 Evaluate the pressure drop in a horizontal 100 ft length of galvanized rectangular duct of 200F dischaiging at 15.0 psia. All we know about the. bed porosity is that 0 < w < 0.6.
dimensions 1 in. x 3 in. for Evaluate the bed poro.i;ity from the above conditions and the data below:
a) an average air flow velocity of 1.5 ft/sec at 100F and 15 psia; Dp = 0.1 ft
b) repeat for a flow velocity of 15 ft/sec. I'/ = 0.05 Ihm/hr ft
3.3 Show that for flow through a slit with a spacing much less than the width, Eq. (3.22) gives p = 0.0075 lbm/ft 3 at 200F, 15 psia.
the friction factor for laminar flow. 3.9 In the production of titanium, rutile (Ti0 2 ) is fluidized with gaseous chlorine, and the
3.4 Refer to Problem 2.10 and calculate the maximum diameter of spherical alumina reaction --
particles that would not be expected to obey Stokes' law.
3.5 A steel ball (radius = 0.291 ft) is dropped through molten slag in order to determine its occurs with the oxygen being continuously removed by reaction with coke particles,
viscosity. The steel's density is twice that of the slag and a terminal velocity of 5 ft/sec is
experimentally determined: Calculate the kinematic viscosity of the slag in ft2 /sec.
The rutile ore has a screen analysis of
3.6 A tower 50 ft high and 20 ft in diam. is packed with spherical pebbles (0.1 ft in diam.).
Gas enters the top of the bed at 1000F and leaves at the same tempemture. The pressure at -90 +SOM 5.0%
-80 + IOOM 6.2%
-100 + 140M 77.4%
-140 + 200M 11.4%
Calculate the possible range of flow of the chlorine at 950C which is needed to fluidize 4
the rutile in a bed 4 ft in diam. Give your results in terms of lb Cl 2 /min, and in terms of velocity,
ft/sec.
ENERGY BALANCE APPLICATIONS
i{)OJDuring the compaction of metal powders into sheet material (powder rolling), as shown IN FLUID FLOW
in the figure below, the entrapped air is expelled from the loose powder. This expulsion occurs
at the line AB. Below AB, the powder is coherent, that is, the particles are locked together, but
abpve AB, the powder is loose, i.e., a normal packed bed.
ttt One of the most powerful analytical tools available for quantitative problems in
engineering is the principle of the conservation of energy. In fluid flow, this
principle is most often applied to open systems in the form of the mechanical energy
balance which is also commonly called Bernoulli's theorem. The approach is
macroscopic, since we examine the entire flow system with an entrance and an exit.
This method is different from the one presented in Chapter 2, where we analyzed
' the microscopic volume element through which fluid flows as the starting point for
solving problems.
_. --hs
4.1 CONSERVATION OF ENERGY
If the velocity of expulsion exceeds the minimum fluidization velocity, the powder d9es
not feed properly into the roll gap and the sheet product is not satisfactory. This, in fact, limits
the production rate of the process.
The conservation of energy can be applied to the flow offtuids such as that depicted
in Fig. 4.1. With the aid of the symbols in Table 4.1, the general energy balance can
be written
The following equation gives the superficial velocity V0 (cm/sec), of gas expulsion from (4.1)
the coherent zone:
2nR 2 n(<X - sin <X cos <X)
In this form, the macroscopic energy balance is written for unsteady-state condi-
Vo= . , tions in which the left side of the equation represents the accumulation of total
<X[2R(l - cos <X) + h,.]
energy in the system. The total energy, 101 , is defined as the sum of internal,
where R = roll radius, cm, n = rolling speed, rev/sec, <X = roll-coherent powder contact angle,
rad, and hs = roll gap, cm. r,-:. .
For copper powder, the angle <X has been found to..be 6,
a) Calculate the rolling speed at which copper powder, with properties given below, will
just begin to be fluidized at the plane AB, for strip thickness 0.05 cm and roll diameter
20cm.
b) Calculate the corresponding strip speed.
c) Calculate the maximum allowable rolling speed for this powder (elutriation conditions).
d) Discuss the effect that order-of-magnitude changes in particle size and roll radius would
have on production rates. Data are given as follows:
4 3
l'/air = 1.8 x 10- poise, Pair = 1.3 x 10- g/cm3, DPcu = 0.004 cm, Pcu = 6.9 g/cm 3
Fig. 4.1 Flow system for the application of the energy balance.
111
Table 4.1 Symbols used in the general energy balance
Examine Eq. (4.2) for meaning; one may be tempted at this point to simply write
!{pA 1 V1 )V12 to represent the kinetic energy entering the system. However, because
Entering the system Leaving the system
the velocity varies in some manner across the cross-sectional area, the integration
Mass flow rate indicated by Eq. (4.2) is in order.
To account for the velocity distribution, the kinetic energy term can be written
Pressure
as:
Volume per unit mass 1/pz
(4.3)
Energy per unit mass
Potential Epi EP2
Kinetic EKl EK2
where /3 1 is defined by
Internal Ui U2
Enthalpy Hi = U1 + P1(1/p1) H2 = U2 + P2(1/p2) (4.4)
Mechanical work done by the system, M
Net energy generation within the system We can also use a similar expression for the rate at which kinetic energy leaves the
from chemical reactions or other
system; so we write the difference between leaving and entering the system
sources, SR
Net heat input, Q i7z2
w AEK = w (2[3 2 -
V/)
2/31 . (4.5)
potential, and kinetic energy. The operator A signifies exit minus entrance quanti- For laminar flow in a circular tube, f3 = 0.5, and for turbulent flow, f3 is nearly
ties, and note that the expression AH = AU + A(P/p) has been employed in the unity, and is usually approximated as such.
usual manner for open systems.
4.1.2 Evaluation of the potential energy terms
The terms Pi/p 1 and P 2 /p 2 , which are parts of H 1 and H 2 , respectively, are
often called the flow work or PV work. They are recognized as the work required Potential energy is defined relative to some arbitrary reference plane for the fluid
to put a unit of mass into the system at "1" in Fig. 4.1 and the work done by the both leaving and entering the system. Therefore, the difference in potential energy
system on a unit mass leaving at 2, respectively. After this brief discussion explain- of the fluid leaving and entering is simply
ing the basis of the problem at hand, namely, the flow offluids, we now proceed to
focus attention on some of the specific terms in Eq. (4.1) for steady-st:,ite conditions,t
WAEp = Wg(z 2 - z1),
,f I namely, or
' '
WAEp = WgAz. (4.6)
and
4.1.3 Mechanical energy balance, Bernoulli's theorem
I 4.1.1 Evaluation of the kinetic energy terms
!
Substituting Eqs. (4.5) and (4.6).into Eq. (4.1) for steady-state conditions yields
I The rate of kinetic energy entering the system through area A 1 (normal to flow) is
I I
given by
AH + ( i7z2 - V/ ) + g Az + M* - Q* = 0. (4.7)
'f I A1 2/32 2/31
1'
I
' W1EK1 =~ [J (p 1v1)vf dAJ. (4.2) Here M* = M/W, and Q* = Q/W. In this form, all the terms are now expressed
0
per unit mass of fluid flowing through the systt?m. Further, if we express H by
t U + P / p, we obtain the overall energy balance form of Bernoulli's theorem:
t For a less specific but still concise presentation of the general energy balance, refer to D. M.
Himmelblau, Basic Principles and Calculations in Chemical Engineering, Prentice-Hall,
Englewood Cliffs, New Jersey, 1967, pages 261-281 and 397-401.
I AU +A (- )+ (-V2-2 - -2/31Vi-2) +
P
p 2/32
g Az + M * - Q* = 0. (4.8)
'lII
L
-..,------.-i-q:-r;nergy-ua1ance-app11cauons-1n-uu10-uow---,--- FrictiOiilosses in straight condiii~rr
11
The equation is also commonly found in its differential form for a differential only due to a pressure difference of P2 - Pi. then Eq. (3.3) gives us the force exerted
segment of the system: by the fluid on the wall.
where c5E1 is the mechanical energy per unit mass lost as frictional conversion into For circular tubes, A = nR 2 and F K = 2nRL(tp 17 2 )f. Thus,
heat. Substituting Eq. (4.10) into (4.9) and expanding d(P/p) into
[(1/p) dP + Pd(l/p)]
Ef = 21(M 17 2
. (4.15)
f p + [-2 Vi-2 J+
P2 dP V2
2/32 - 2/31 g Llz
* _
+ M + Ef - 0. (4. i 1)
a~ a ~ate of 1000 ft 3/min. What theoretical horsepower is required of the fan if the
air discharges at 750 mm Hg?
P1 Solution. The problem does not state whether the duct is horizontal, vertical, or
Actually, Bernoulli's equation only applies to the flow of ideal fluids, where at an angle; thus, t~e potential energy change of the gas as it passes through the
M* = Er = 0, but Eq. (4.11) having been used so frequently in engineering, has system cannot be defined. However, since the density of a gas is quite small, we
taken on the connotation of Bernoulli's equation as well. Also note that it is may safely ignore the potential energy term in Eq. (4.11). Consider the system
below.
written in terms of unit mass of material flowing.
~-&=-.F-a-n--~-=-------~--~!~ ..
..______...
4.2 FRICTION LOSSES IN STRAIGHT CONDUITS
Due to the utility of Eq. (4.11) in engineering calculations, considerable effort has ~~ _ ~
been made to develop methods for estimating the friction loss E1 in various flow ~ 180ft.,~--------tt..;1 I
systems. (1) (l)
Ifwe consider the friction loss in a straight conduit with steady flow of constant- ~
density fluid, we can determine E1 from the friction factor f. If a conduit has an . With the fan included within the system, Bernoulli's equation is written to
arbitrary but constant cross-sectional area A and a length L, and if the fluid flows Include M*.
-- -----------~..--.;---------.,.
P2
Now = 0.106.
2(1 x 1) ft 2 For straight conduits, Eq. (4.15) gives the friction loss Ef. However, in most
De = (i + l) ft = 0.80 ft. flow systems there are fittings, bends, changes in cross sections, valves, etc. There-
fore additional resistances must be included in EJ for the best application of
At 1 atm and 60F, the density of air is 0.0763 lbm/ft 3 . Therefore, Eq. (4.11). The following sections review some of the methods used to calculate
such resistances.
~ 750 mm Hg 0.0763 lbm _ 3
p = 760 mm Hg ft 3 - 0.0753 lbm/ft .
4.3 ENLARGEMENT AND CONTRACTION
(Why is the approximate sign rather than the equal sign employed here?) Also, The friction loss associated with a sudden enlargement or a sudden contraction
Y/ = 0.043 lbm/hr ft, (Fig. 4.2) is calculated by using a friction-loss factor ef. The friction loss for the
_ 1000 ft 3 60 min particular geometry that upsets flow in conduits is evaluated in the following manner :
iI
V2 = -m--n---1-i~f-t2-i---h-r- = 90,000 ft/hr.
1 1 (4.16)
The Reynolds number is now determined: where Vis the velocity of fluid in the smaller cross section. (Since other conven-
e ;-.eV2-;.
= 1.26
From Fig. 3.2, f = 0.0042 (smooth). Since
Y/ /
___.,.---""
J (0.8)(90,000)(0.0753)
x 10 5 .
(0.043)
tions are also used, the reader should always be certain of which velocity is chosen
to define ef.) By analogy with Eq. (4.15), ef corresponds to a friction factor with
built-in geometry effects. Hence, ef depends on a Reynolds number and geo- a
metric ratio of the system. We present values of ef for enlargements and contrac-
tions in circular cross sections in Figs. 4.3 and 4.4, respectively .. For both these
figures, D is the smaller diameter, Re is calculated using the same D, and L is the
2
length of the. smaller pipe considered.
Ef L )-
= 2 ( De V f,
f't I
I I
the energy balance can be written as I
I
I ~L
M* + [ 2~ 2 + 2f(~J]v 2
= 0. I AIF
I
! I
{I) (2)
For /3 2 ~ 1 (turbulent flow),
i I
I I
1127"
M* = -[ ~ + (2)(0.0042)( ~.~~) }9 x 4 2
10 )
2
ft /hr
2 I
I
I Ii St---
I
I
I I
1.94 x 10 10 ft 2 1 hr 2 sec 2 -lbf I I
',,
Fig. 4.2 Sudden contraction an~ enlargement.
~ hr 2 3600 2 sec 2 32.2 ft-lbm
The results for ef in the case of enlargements apply equally welJ to all exit
t By stating that P 2 = 750 mm Hg, we ignore the exit loss which we shall discuss in Section 4.3. shapes (except for gradual expansions discussed in the following paragraphs), as
- - - - - - - - - - - - - - - - -- ----- - - - - - - - - - - - - - - - - - - - - - - ,
*~ 1.4
1.2
Turbulent
Re= oo
0.6 = 10,000
= 3,000
0.4
0.2
0.0
-0.2
,,\i -0.4
-0.6
-0.8
0.0 0.2 0.4 0.6 0.8 - 1.0
A1/A1
Fig. 4.3 Friction-loss factor for sudden expansion. (From W . M. Kays and A. L. London, -0.2
Trans. ASME 74, 1179 (1952).)
-0.4
can be seen by viewing the flow in Fig. 4.2. The extent of the region of vortices
after the expansion does not depend on whether the corners are rounded. However, -0.6
e1 for contraction (entrances) can be modified significantly. Figure 4.5 shows the .,, .
ratio of e1 for various inlet shapes to that for a sharp-edged entry. -0.8
In flow through gradual enlargements, the energy losses are significantly
0.0 0.2 .0.4 0.6 0.8
reduced due to the elimination of the vortices shown in Fig. 4.2. Experiments on
tapered enlargements show that the friction-loss factor e1 depends on both the
taper angle f3 and the area ratio Ai/A 2 (Fig. 4.6). In reality, e1 should also depend Fig. 4.4 Friction-loss factor for sudden contraction. (From Kays and London, ibid.)
on a Reynolds number, but this is not given in the figure, since, in turbulent flow,
the dependence of e1 on the Reynolds number is small, as can be seen in Figs. 4.3
and 4.4. Consequently, data presented for e1 are usually assumed to apply to all
Reynolds numbers in the regime of turbulent flow.
In gradual contractions, the converging boundaries have a tendency to steady
the flow, and if the included angle of convergence is 30 or less, the energy loss factor
can be approximated as 0.05-0.10. Again, when using these data, one must realize
1~
.I that they apply only to turbulent flow .
"
4.4 FLOW THROUGH VALVES AND FITTINGS
'
To evaluate flow through valves and fittings, we usually assign an equivalent length
to the fixture such that the pressure drop is given by Eq. (4.15), where Lis replaced
by Le. Therefore
d/2
E = 2/Le j72 (4.17)
'f D ,
Fig. 4.5 Entrance-loss coefficients.
4.5 FLOW THROUGH SMOOTH BENDS AND COILS
-~
A,
In curved pipes, the friction loss may rise considerably above the values for straight
{3 = 60
lengths of pipe, and the transition Reynolds number may be much higher than
{3 = 90
2100. The maximum Reynolds number for laminar flow, or the critical Reynolds
number, Re"' is given as a function of the coil diameter De and pipe diameter D :1
D)o.32 DC
Rec = 20,000 De ( , 15 < D < 860. (4.18)
Fig. 4.6 Friction-loss factor for turbulent flow through gradual enlariements. (From ~ 0.25
Steam-Its Generation and Use, Babcock & Wilcox, New York, 1963, page 8-17.)
where f is evaluated for Re with the pipe diameter equal to the fitting diameter, 0.20
and Le/D for turbulent flow of various fittings is given in Table 4.2. Thus, when I
Laminar
evaluating the friction loss through a system with fittings, the sum of the equivalent 0.15
flow
lengths of all the fittings is added to the length of the pipe; E1 is then determined by region
using Eq. (4.17). 0.10
For example, if a pipe network contains 20 ft of i-in. pipe (I.D. = 0.824 in.), Turbulent
flow
two elbows (90, medium radius), and a wide open gate valve, then the L/D ratio 0.05 region
to be substituted into Eq. (4.15) would be
0
2000 4000 6000 8000
Rec
20 Fig: 4.7 Transition Reynolds number for flow through smooth curved pipe or coil. (Eq. 4.18
= 0.824/12 + (2)(26) + 7 = 351. from H. Ito, Trans. ASME 82, Series D, 123-34 (1959).)
The friction loss for laminar flow in a curved pipe can be expressed in terms
Table 4.2 Equivalent length for various fixtures, turbulent flow*
of the ratio of the friction factor for the curved pipe to that for the same length of
Fitting Le/D Fitting
straight pipe. This' ratio is a function of the Dean number, Re(D/Dc) 1 12 , as shown
Le/D
in Fig. 4.8. Note by referring to Fig. 4.8 that the effect of curvature is negligible for
45 elbow . . .. . .. . ... .. . . . . .. . . 15 Tee (as el, entering branch) 90 Re(D/Dc) 1 12 < 10.
[ __:.:::;---90 elbow, standard radius.. 31 Couplings, unions ......... Negligible We can compute the friction loss for turbulent flow by using Eq. (4.15) and
- 90 elbow, medium radius... 26 Gate valve, open............ 7 applying the ratio of curved-to-straight pipe friction factors, fc/fs, given by 2
90 elbow, long sweep......... 20 Gate valve, t closed......... 40
90 square elbow............... 65 Gate valve, t closed......... 190 !c (4.19)
180 close return bend...... 75 ~ Gate valve, l closed......... 840 ls
Swing check-valve, open... 77 1
:> Globe valve, open............ 340 for Re(D/Dc) 2 > 6.
Tee (as el, entering run)...... 65 Angle valve, open............ 170
L.
------ - - - - - - - - - - - - - - - - - - - - . --------~:11-urrr- ........,.n.'3Ur11E::T1.. ~111:--n:;;,-----
From Fig. 4.7, the flow is shown to be turbulent, and so we apply Eq. (4.19):
Since
Fig. 4.8 Effect of curvature on fin laminar flow. (From Steam-Its Generation and Use,
Babcock & Wilcox, New York, 1963, page 8-17.)
p y2 = Re2(_D1J) 2 _ol = 3.44 x 1010 ~2
ft hr
Example 4.2 An induction melting furnace has a 14-turn coil of! in. l.D. copper-
tubing wound with a coil diameter of 16 in. For adequate cooling, the manu- l1P = (2)(0.0088) 14n x 16 in. 3.44 x 10 1.0 lbm I hr 2 sec 2 lb1 1 ft 2
facturer's specifications call for water at 4 gpm. Determine the water pressure ! in. ft hr 2 2 2
3600 sec 32.2 ft lbm 144 in 2
that should be available in the water supply line in order to deliver the 4 gpm
through the coil if the water discharges at 1 atm.
Solution. For this situation, Bernoulli's e_quation reduces to:
Thus. the line pressure must be at least 14.2 psig.
P2 - P1 L1P
EJ = = _,
p p
4.6 FLOW MEASUREMENT
neglecting the potential energy contribution. The task at hand, then, is to evaluate
the friction loss E1 : Efficient operation apd control of engineering processes and experimental set-ups
often require information concerning the quantities of flowing fluids. For measur-
DVp 4W
Re=--=-- ing flow in closed conduits, there is a variety of measuring devices available; for
Y/ nDYf example, velocity meters, head meters, or area meters. Although all of them can
where Wis the mass-flow rate (lbm/hr). be used to determine the mass-flow rate, which is the information usually needed,
none can measure the flow directly, so that, particularly in the case of compressible
Re = 4 4 gal 8.33 lbm 60 min 24 ft hr fluids, the meters require careful calibration.
n min I gal . I hr I ft 2.42 lbm
4.6.1 Velocity meters
4
= 2.52 x 10 . A commonly encountered velocity meter is the l'litot-static tube which measures
local velocities. The pitot tube consists essentially of a tube with an open end
D) 112 ( 1/2) 112 facing the stream, as shown in Fig. 4.9. The velocity of the fluid along the streamline
(D c
= l6 = 0.177.
x-y decreases to zero at y which corresponds to the tip of the pitot-tube opening,
,---------~--- ----------------------
and is called the stagnation point. Applying Bernoulli's equation between planes (1) run of at least 50 diameters without obstructions, the ratio of vmax (the velocity
and (2), we have measured at the center line) to V (the average velocity) can be found by relating the
ratio to the Reynolds number. For laminar flow, of course,
P2 + O = P1 + vi , (4.20)
p p 2
0 < Re < 2.1 x 103 , (4.22)
where P1 denotes the upstream pressure, and v1 the approach velocity. The
pressure P2 , which is called the stagnation pressure, is made up of two parts: and for a large region of the turbulent range, namely, 104 < Re < 10 7 ,
(1) the pressure to be measured if the fluid were not moving-the static pressure,
and (2) the pressure resulting from the sudden cessation of the streamline's kinetic ; I
j7 = 0.62 + 0.04 log( D VmaxP) (4.23)
energy-the velocity pressure. In other words, vmax 11
Jn the range 2.1 x 103 < Re < 104 , a traverse must be made because V/Vmax is
not predictable.
For gases at low flow rates ( < 200 ft/sec) and isothermal flow conditions, we
which is, in fact, a rearrangement of Eq. (4.20). The construction of the pitot tube
may use Eq. (4.21) as it is. However, if there is a substantial temperature rise (high
in Fig. 4.9 shows the static holes; at these locations, static pressure is measured,
velocity), then an energy balance must be made between the upstream point, at
while the pressure differential P2 - P 1 is measured by the complete assembly.
which a temperature is measured, and the tip of the pitot tube. For this case,
Thus, after including a calibration constant CP which takes care of nonideal
Bernoulli's equation should be written
behavior in the system, we rewrite Eq. (4.20) to yield the velocity of the fluid:
CP, which is sometimes called the pitot-tube coefficient, usually has a value between If the fluid is a compressible gas, we may assume that the flow is adiabatic
0.98 and 1.00. and the gas ideal; thus the equation of state is
p(~r = constant,
~
(4.25)
---------t-
I '~~ D '
where y is the ratio of the heat capacity at constant pressure to that at constant
~
~
. t
8 static holes
t3v volume. Inte_grating Eq. (4.24) by using Eq. (4.25) yields
0.002 to 0.04 in.
(4.26)
Impact opening= 0.4D
Low-pressure
side,P1 Example 4.3 A pitot-static tube is installed with its impact opening along the
center line of a long pipe with an I.D. of 1 ft. Air at 150F and 12 psig passes
High-
pressure through the pipe; the barometric pressure is 745 mm Hg. The pressure difference
side, P 2 - measured by the pitot-static tube is 0.42 in. water. Calculate the total flow rate
Fig. 4.9 Pitot-static tube and recommended dimensional relationships. in the pipe (lbm/hr). '
Solution. Here we apply Eq. (4.21) and, in the absence of data, assume that CP = 1.
Because the pitot tube measures only local velocities, a traverse of the conduit The absolute pressure in the pipe where the meas1.rn~ment is made is
must be made in order to obtain the complete profile, and from it the average
velocity V. To obtain the density p, we usually measure the temperature upstream 12 745
P1 = - - atm + - atm = 1.80 atm.
from the probe tip. In the flow in a straight circular pipe at points preceded by a 14.7 760
)
L
-----------------------------~;---------
T_hen we calculate the gas density: the flow line, causing a local increase in the velocity of the fluid and a corresponding
I lb-mol 1-80 atm 492R 28.8 lb decrease in the pressure head.* The simplest is perhaps the orifice plate illustrated
1 atm 610R l lb-m~l = O.ll 6 lbm/ft .
3
p = 359 ft 3 in Fig. 4.10; however, the venturi meter and flow nozzle (Fi-gs. 4.11 and 4.12) are
based on the same principle, and the same group of equations apply to all of them.
We convert the measured pressure difference into more convenient units
'
p _ p = 0.42 in. H20 5.20 lbf/ft 2 32.2 lbm-ft _
2 1 . H O
1 m. 2 ~ sec 2 - :r
Now we utilize Eq. (4.21) to determine Vmax:
_ 2
lb - 70.4 lbm/ft sec .
-
F low
2 ft 2
(70.4)-2 = 34.8 ft/sec.
0 116 1 sec
(At this velocity and pressure drop, the use of the more complicated expression in
Eq. (4.26) is not necessary.) Fig. 4.11 Venturi meter.
To find the average velocity, we utilize Eq. (4.23):
- Flow
the mechanical energy balance (for incompressible fluids) with f3 = 1 is
P2 - P1
p
Vl Vl
+1-2=0.
Also, for incompressible fl~ids (by continuity),
(4.27)
(4.28)
Fig. 4.10 Orifice meter. *We shall discuss the term head in Chapter 5.
- - - - - - - - F ' l o w measurement --129
so that when Eq. (4.27) is solved for V2 in terms of the difference in pressure between
the taps, we obtain Expansion factors for gases are different for each type of head meter; these
re summarized in Table 4.3.
2(P1 - P2 ) a For convenience, Fig. 4.13 shows values of Y for each of the above met~r types.
(4.29)
(~:rr
. urds , y is taken to be unity ' as it is also for gases when the pressure difference
for 11q
1
p( - is small compared with the total pressure.
This is the theoretical velocity at the vena contracta, which disregards frictional
energy losses and can never be achieved in practice. Therefore we utilize a discharge
coefficient CD which accounts for such losses and an additionalgeometric factor. The
largest pressure drop is measured at the vena contracta, but it is more convenient
to relate the velocity at the orifice plate V0 to the pressure drop, (P1 - P ), and at
2
the same time use the diameter of the plate opening D0 instead of the diameter of
the vena contracta. Thus we 'get the final form of the expression
2 (P1 - P2 )
p (1 - B 4 ) ' (4.30) Nozzles
Venturis
in which B =D /Di. or
0
}
o.8 L----'----'------'-----.l-----'-,-----:--'-:-,-----~-~:::--!
(4.31) 0 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0
.16
1-r
where K is the flow coefficient, defined as CD/~ k
Equations (4.30) and (4.31) also apply to venturi and nozzle meters as well as Fig. 4.13 Expansion factors for gases flowing through various head meters. (From Perry,
to orifice plates, but, as will be discussed below, CD and K are quite different. ibid., page 403.)
In addition, for both venturi and nozzle meters, as shown in Figs. 4.11 and 4.12,
the minimum flow contraction corresponds to the position where P 2 is measured;
For the orifice plate, we have to choose: (1) whether to use a square-edged
hence D 2 = D 0 , and V2 = V0
hole or one with a 45 -taper and a knife-edge inlet, and (2) where to locate the
For liquids (incompressible fluids), the mass flow rate is
pres~ure taps with respect to the plate. Figure 4.14 illustrates the effect of various
(4.32) locations of taps on the flow coefficient, and shows how the stream is narrowed as
it leaves the orifice plate, reaching a minimum at the vena contracta. In most cases,
For gases, we can modify this equation to a more general form, utilizing the the upstream tap is located from one to two pipe diameters ahead of the plate, and
expansion factor Y,
the downstream tap either one half of a pipe diameter downstream or exactly at the
(4.33) vena contracta which would have to be found experimentally. As specified in the
legend, the da~a of Fig. 4.14 apply specifically to square-edged circular orifices.
Table 4.3 Expansion factors Y for flow meters* (Jenerally, such data depend to some extent on the exact design of the orifice plate
Meter type y and on the location of the pressure taps so that, to be safe, one should rely on
manufacturer's information and/or self-calibration. However, there are data
Venturi and nozzle j y(l-r 1 - 1 iY)(l-B 4 ) available in the general literature for various designs that may be consulted. 3 ' 4 ' 5
(}' - 1)(1 - r)(r-2 1r - B 4 )
P. S. Barna, Fluid Mechanics for Engineers, second edit~on, Buttersworth, Washington D.C.,
3
K 0.95 The design of the venturi meter is such that a gradual restriction in flow area
precedes the throat, which is a short, straight section; then the flow area gradually
0.90
/I
ff &
oO
returns to the original area. On the upstream side, the included angle of conver-
nce is about 25, and on the downstream side, we employ an included angle of
~~vergence between 5 and 7. A venturi tube is nearly frictionless under turbulent
flow conditions, so that the typical values of C1? are between 0.98 and 1.0. H.owe~er,
0
o"
//
in laminar flow, Cn drops rapidly with decreasing Reynolds number and cahbrat10n
0.85 . data should be consulted.* For venturi meters, the permanent pressure drop is
I
I
much lower than for orifice plates and may be approximated as ten percent of the
I
I
measured differential.
I Nozzles are similar to orifices in some respects, but are designed so that the
I
0.80 ; discharge is preceded by a smooth contracting passage. As a result, nozzles have
less eddying upstream and the measured pressure drop corresponds more closely
to that for a venturi rather than an orifice meter. However, downstream from the
nozzle discharge, the flow behavior is more similar to that of an orifice; thus the
permanent pressure drop corresponds more closely to the orifice than to the
venturi.
4.6.3 Area meters
0.70 These meters are also based on the principle of placing a restriction on the flowing
stream, creating a pressure drop and a corresponding change in flow velocity
through the restricted flow area. However, in this case the pressure drop stays
constant and the flow area changes as the velocity changes, rather than vice versa.
0.65
The most common type of area meter-called a rotameter-is illustrated in
Fig. 4.15. The flow is read by measuring the height of a float in the slightly tapered
o.6o0l~~~~~~~~~===:=~~-'--
column.
A force balance applied to the float determines the equilibrium position.
o.5 .2 3 4 5
Downstream pressure-tap location in pipe diameters When a fluid of density p moves past the float and maintains it in suspension, we
can use the same force balance that was used several times in Chapters 2 and 3 for
Fig. 4.14 Effect of tap locations on the flow coefficient. Applies to square-edged circular particle dynamics. The net buoyant weight of the float is balanced by the upward
orifices with Re (at the orifice) greater than 30,000. (From Perry, ibid., page 405.)
force created by th~ moving fluid. This is expressed as
mf
The square-edged orifice plate is most commonly used, but we use the sharp- (pf - p)-g = FK, (4.35)
edged plate to ensure reproducible coefficients when the plates must be often PJ
replaced, and recalibration is to be avoided each time. This is a problem, 'for where mf is the mass of the float, and PJ is the float density.
example, in systems where the gas is dirty and the velocities h'igh, which results in For a given meter through which a given fluid flows, the left side of Eq. (4.35)
rapid erosion of the opening in the plate. is constant and independent of flow rate. Accordingly, F K is constant when the
Orifice plates are the simplest and cheapest types of head meters, but they also float is at equilibrium, and, if the flow rate changes, the float counters the effect of
cause the largest permanent pressure drop in the system. This refers to the fact this change by taking on a new equilibrium position. For example, if the float is at
that. the pressure difference, P1 - P2 , across the orifice is not entirely lost, and a some equilibrium position corresponding to some.mass flow rate and then the mass
partial pressure recovery is experienced downstream. The results of experiments flow rate increases, FK becomes larger, and the float rises. However, as the float
indicate that the permanent loss of pressure for orifice plates is given by
~Ploss = (1 - B 2)(P1 - P2). (4.34) *For example, see Perry, ibid., page 407.
-----~- -- - - - --~-----r-------
rises, the tapered tube presents a larger cross-sectional area for flow, and the When we apply a dimensional analysis to a float of a given geometry, the
velocity of the fluid between the float and the tube wall decreases, so that a new following functional relationship between dimensionless groups evolves 6
equilibrium position is eventually reached, where F K returns to the value expressed
by Eq. (4.35). W _ !(JmJgp(l - PIPJ) , D,), (4.36)
The variety in designs of rotameters is so great that there does not exist one DJjmJgp(l - p/pJ) 11 DJ
relationship valid for all types of rotameters to describe how the mass flow rate
where W = mass flow rate, DJ = characteristic diameter offloat, and D, = diameter
varies with height. However, the manufacturers usually supply calibration data
for their devices, each set of data being appropriate for a specific fluid. Thu~, if a of the tube. .
The ratio D,/ DJ, of course, is directly related to the meter readmg h, so that
gas mixture such as He plus 10 % 0 2 is being passed through a rotameter calibrated
Eq. (4.36) does show the general form W = W(h). Equation (4.36). is impo~tant
for He alone, then the user is fooling himself. The rotameter must be recalibrated
because it indicates how the same set of cu~ves g~nerated to ~~ Its functional
for the He plus 10 %0 2 mixture, or a dimensional analysis of the system that would
relationship can be used for all fluids. Thus, If one mten?s to utilize a rotameter
clearly indicate how the physical parameters interact must be carried out.
for many different fluids (for example, as a laboratory item), one should know
enough characteristics of the rotameter to be able to make it completely versatile.
Flow indicator
-
(a)
metering-tube
Flow
In
Wobble disk
(b) Radial partition between
-------Float inlet and outlet ports
.----Thermometer
.,.---Gas outlet on
,,...-.,........_ back of meter
~""---'-Gas inlet on
back of meter
Water level
Gas-inlet slot
to bucket
Gas-measuring
rotor
(c)
Float stop Fig. 4.16 Various flow totalizers. (a) Rotary vane meter,(~) rotating disk meter, and (c) liquid-
sealed gas meter. In each case, the shaft feeds a mechanicarcounter.
In 6
W. L. McCabe and J. C. Smith, Unit Operations of Chemical Engineering, McGraw-Hill,
Fig. 4.15 Rotameter. New York, 1956, pages 120-121.
~------------e.,----r-r----------------~--~----------------
Now consider1he case when the mass flow rate of the incoming stream is
reduced so that the liquid level gradually diminishes. Then the velocity of the exit
stream changes with time, but the velocity at any instant is still given by Eq. (4.41),
where h is the liquid level at that instant. Finally, if the vessel contains a liquid and
t7z2 + g Az + EI = 0. (4.37)
2132
Stopper
The friction loss EI for this system is entirely due to the contraction nozzl~;and
we can evaluate it by the method applied in Section 4.3 : ''---------Nozzle
no liquid enters at the top, and then a bottom plug is removed, Eq. (4.41) still
applies, but h must be recognized to be a function of time. This is the situation that . solution .
applies to ladles; Fig. 4.18 shows a typical ladle design for molten steel. . Equation (4.46) may be utilized if it is reasonable to treat CD as a const~nt. Smee
The mass flow rate discharging from the ladle with a nozzle of area AN is given - e and [3 2 both depend on the Reynolds n~mber through ~he n~zzle, CD is. actuall_y
by :function of the height. ~et.us first examme the manner m which CD v~ne.s. This
involves calculating the fnct10n-loss factor for flow through a contract10n, hence,
(4.43) we use Figs. 4.4 and 4.5.
dw = nDf dh .
Re~ 0, e1 ~ 0.8, /3 2 ~ t (laminar),
-p4 (4.46)
CD ~ [2.8]- 112 ~ 0.60.
Thus we obtain the time t1 to completely empty the ladle by integrating Eq. (4.45): I-in. level
Example 4.4 A ladle having an l.D. of 3 ft with a capacity height of 4 ft has a Calculations at all levels show that CD is constant at 0.96 except for the very
nozzle that tapers (45) to a 3-in. exit diam.: a) Calculate the time it takes to empty last bit. Hence, it is a good assumption to treat CD as a constant.
the ladle if it is filled with Al-7% Si alloy at 1300F. b) Calculate the discharge J, '----0 fe_d ""~'O
rate (lbm/sec) (1) initially, and (2) when the ladle is 75 % empty.. ~e.sowr.. J~ Ql~\'f\lo nD2 (ho)112
a) t = - - -
Data: '1 = 2.75 cP = 6.59 lbm/ft hr,',f \' 1
2ANCD 2g
p = 150 lbm/ft\
' ) ~ = 74.9 sec.
11
9 4
= (n)(
2
) (
2n(i) (i)(0.96) 2 x 32.2
= 113.3 lbm/sec.
Discharge rate at 75 % empty (1-ft level)
PROBLEMS
4.4 A venturi meter is installed in an air duct of circqlar cross section 1.5 ft in diameter which
4.1 Hot-rolled steel sheet is quenched by passing under nine water sprays as depicted below. carries up to a maximum of 2500 ft 3 /min (STP) of air at 80F and 1 atm.
,II Each spray requires 15 gpm of water at 70F, and the pressure drop across each nozzle at this
flow rate is 25 psi. Calculate the theoretical horsepower required by the pump.
a) Determine the maximum pressure drop that a manometer must be able to handle, i.e.,
what range of pressure drops will be encountered? Express results in in. of water.
I ~3ft---~1 b) Instead of the venturi meter, an orifice meter is prop_osed and. the maximum pressure drop
to be measured is 2 in. of water. Calculate what diameter of sharp edged orifice should be
installed in order to obtain the full scale reading at maximum flow.
c) Estimate the permanent pressure drop for the devices in parts (a) and (b).
I-in. I.D. piping d) If the air is supplied by a blower operating at 50% efficiency, what is the power consump-
e/ D= 0.0004 tion associated with each installation?
4:s""fiQ
practice, when metal is poured from ladles, the nozzles erode. The erosion can be
:i Connection (Tee)
tetd~d to the quantity of metal that passes through by this simple expression:
DN - DZ= kW,
to 3-in. line
in which DZ = initial nozzle
---=-
diameter, DN = nozzle diameter during pouring that varies,
sump W = mass of metal passed through, and k = proportionality constant.
Develop an expression for the time to empty a ladle that accounts for nozzle erosion.
4.6 In order to slush cast seamless stainless-steel pipe, a pressure pouring technique is utilized,
as depicted below.
Atmospheric pressure
4.2 A fan draws air at rest and sends it through a straight duct 500 ft long. The diam. of the
duct is 2 ft, and a pitot-static tube is installed with its impact opening along the center line.
The air enters at 80F and 1 atm and discharges at 1.2 atm. Calculate the theoretical work
'
(in ft-lb1 /lbm) of the fan if the pitot-static tube measures a pressure difference of 1.0 in. of water.
~~ -~Pitot-statictube ~
.. 1.0 Fan JI ~ I2
atm ~,'"' ..__ _ _ _2_5_0_f_t--SOO:
4
~ oim
4.3 Compressed air at 100 psia and l00F flows through an orifice plate meter installed in a
3 in. l.D. pipe. The orifice has a 1.00 in. hole and the downstream pressure tap location is .-
The mold must be filled.rapidly up to level Li so that no solidification takes place until the
entire mold is filled. Determine an expression that we can use to give the time it takes to fill
1.5 in. from the plate. When the manometer reading is 14.l in. Hg, the mold only. Consider the mold to be between L 1 and Li and open to atmospheric pressure
a) What is the flow rate of air (lbm/hr)? at the top. Neglect tlie change of metal height L 3 in the ladle; you may also neglect the friction
loss associated with the supply tube and mold tube walls, but the entrance loss may not be
b) What is the permanent pressure drop?
neglected.
l
~-----------------------------------------------~
4.7 Calculate the time to fill the mold, as depicted below, with molten metal ifthe metal level
at plane A is maintained constant and the time to fill the runner system (entering piping) is Inside dimensions Diameter, ft Height or length, ft
ignored.
Ladle 3.0 3.0
'1I Data (all may be taken as constant): Mold
Tube
2.12
0.33
3.0
4.0
1'/ = 4 lbm/hr-ft
P = 400 lbm/ft 3
The tube extends 2 ft below the initial melt surface. Density of the steel is 0.272 lbm/in 3 and its
f = 0.0025 (runner) viscosity is 6 cP.
ej(contraction) = 0.1
. (
Open to the atmosphere
1 ej(enlargement for levels below B) = 0
ej(enlargement for levels above B) = 1.0
L/D (90 turn) = 25 -."
/3 = 1.0
Pressurized
chamber
latm
.----......1 '
4.9 In planning continuous casting, we use fluid flow analysis. Consider the illustrated con-
figuration of the equipment, which includes in-line vacuum degassing.
a) Determine the tundish and degasser nozzle sizes which are necessary to operate the system
Dia.0.lft
at a rate of 25 t/hr per strand. Suppose that for operational reasons it is desirable to
maintain tundish and degasser bath depths of 30 and 72 in., respectively.
b) If only t-in. diameter degasser nozzles are available, how would their use affect the casting
operation?
Inside dimensions:
I
Ladle
5
FLOW AND VACUUM PRODUCTION
~
~
l
.' ': Tundish
f
18" The metallurgist must know not only how to measure flow, but also how to produce
1 it. Some methods of producing flow and vacuum, such as gravity flow and suction
lift, are well known, but there are many less obvious types of fans, pumps, and
flow devices which are of particular interest, since they are often required and used
in metallurgical laboratories and plants. In this chapter, we will consider the
fundamentals of the behavior and application of such devices, and also the behavior
of high-velocity gas jets, which are playing an ever-increasing role in metallurgical
processes.
Outlet
I:
I I'
i
I
I
Piston
motion
i,
I
I
are called heads, h, and so
!
1.'
(5.3)
(b) Rotary gear pump
Since Mp is defined in terms of the mass flow rate, then the power required by the
,fa Fig. 5.1 Positive-displacement pumps. In pump (a), reversal of the piston motion results in
,1:1
I~ reversal of the valve positions, but in continuation of flow in the outlet. pump from a motor (brake horsepower) is given by
f
Ii WMp WL'.lh
11'
~ ; PB = 5sO = r (horsepower), (5.4)
(: Since the flow rate is determined by the size and speed of these pumps, the 550
;: ~ major specification problem is to calculate the power which the pump requires to where W is the mass flow rate in lb/sec and the Llh is given in ft-lbf/ lbm. The
I I
'rJ carry out a specific job. This is done by applying Bernoulli's equation (4.11) to the power absorbed by the fluid (fluid horsepower) is
K 41 system. .Consider that the pump and the fluid being pumped (Fig. 5.2) are the
l'
!, WL'.lh
system, and write Bernoulli's equation from the suction point (3) to the discharge
l
1)
~
I and, therefore,
(5.1)
r =Pf. (5.6)
PB
If the frictional loss within the pump itself is accounted for by a mechanical
efficiency r, then the work done by the pump is On many occasions, the flow rate q is given in gallons per minute (gpm) and the
liquid density in lb/ft3. In such cases, this expression is convenient:
(5.2) (5.7)
L
In pumping liquids, there is a limit on the net positive suction-head that can make a theoretical analysis, giving the relationships between the head devel-
canoot be exceeded. This is set by the fact that if the dynamic pressure of the oped !lh, and the volumetric flow rate Q:
liquid (P + p V 2 /2/3) falls below the vapor pressure of the liquid Py, then the liquid !lh = VP(Vp - Q/Ap tan /J), (5.11)
vaporizes, and no liquid is drawn into the pump. This phenomenon is called
cavitation. For example, consider the liquid just before it enters the pump at where ~ = peripheral velocity, AP = cross-sectional area of the volute channel
plane (2) in Fig. 5.2. In order to avoid cavitation, around the periphery, Q = flow rate, ft 3 /sec, and f3 = angle between the blade
and a tangent to the hub, as shown in Fig. 5.4.
t7z2+ -
P2 ) Py
(- >- (5.8)
2/32 p p {3>90
The pump lifts the liquid from the reservoir through a pipe, and Bernoulli's
equation, written between the reservoir surface (3), which is open to the atmosphere,
and the suction entrance of the pump (2), assuming V3 = 0, is
(
P
p
2 P
-p 3) -
+ 2t7z2
/3 + g(z 2 - z3) + E1 = 0. (5.9)
2
{3< 90
From Eqs. (5.8) and (5.9), we obtain the following requirement in terms of the
height of lift and the pressure at the reservoir:
Q
P Py
- 3 - g(z 2 - z 3) - E1 > - Fig. 5.4 Schematic diagram showing the effect of vane configuration on the shape of the
p p
theoretical head versus flow-rate curve.
Therefore, the net positive suction-head defined below must indeed be positive:
If f3 is less than Q0, then the slope of the theoretical !lh-Q curve is negative;
hnps __ (Pp _Py)p -
3
g(z2 - z 3) - E1 > 0. (5.10) if f3 equals 90, then the line is horizontal; and if f3 is greater than 90, then the
slope is positive. Unfortunately, there are losses due to circulatory flow, fluid
5.1.2 Centrifugal pumps friction, shock, and mechanical friction, none of which is predictable, except by
In centrifugal pumps (Fig. 5.3), the fluid enters axially at the suction connection, experience. Figure 5.5 illustrates the effect of these losses on a pump with f3 less
accelerates radially out along the blades of the impeller, collects in the volute, and than 90.
leaves the pump tangentially at a high velocity. For an ideal centrifugal pump, we The efficiency of centrifugal pumps, which is given by Eq. (5.6), is a maximum
at the design conditions, falling off on either side of them (Fig. 5.6). Pumps are best
Outlet
(I)
Shaft a----lnlet
*:>.
120 Next we write Bernoulli's equation from planes (3) to (1). In this instance,
"<I)c: Eq. (5.1) applies, with L = z 1 - z 3 and V3 = 0, /3 1 ~ 1, and p 1 = p 3 = p.
c;
E p - p) j7,2
"
'O
c:
"
80
70
( p + T + gL + E
1 3
1 = - M*,
60
= 20 I~1
3 2
50 4.0 .... ft 144 in = 46 .2 ft-lb1 .
40
30
3.5 "~"
0
0.
in 62.4 lbm ft 2 lbm
3.0
0
2.5
..c: At 100 gpm, the velocity in the I-in. pipe is 32.1 ft/sec and 3.56 ft/sec in the 3-in.
"
0
2.0
co""'" pipe. Thus,
2 2
0 30.0 60.0 90.0 120.0 150.0 180.0 1.5 V/ = 2
(32.1 ) ft sec lb1 = .0 ft-lb1 .
16
Capacity, gal/min 2 2 sec 2 32.2 lbm-ft lbm
Fig. 5.6 Characteristic curves for a centrifugal pump.
Now we formulate E1 :
described in terms of the head-flow rate characteristic curve. The head is usually
given in terms of feet of the fluid flowing through the pump, and indicates the E1 = 1 V3_
2e1 -2 2 (L)-
+ 2f D V3_2
2
+ 2f D Vi2 (1)-
height of a column of the liquid which could be pulled up through a frictionless entrance
loss
3-in. pipe I-in. pipe
loss loss
pipe to the pump itself. In practice, of course, the friction loss and any fitting or
meter losses would have to be subtracted from this height. From Figs. 4.4 and 4.5, we obtain
e1 = (2)(0.4) = 0.8.
Example 5.1 A strip of metal emerging from a set of rolls is to be cooled by means
The respei:;tive Reynolds numbers and friction factors are:
ofa spray of water. If 100 gpm are required, and the pressure drop across the spray
nozzle is 2o' psi at this flow rate, is it possible to install the pump, as depicted below, Re f
if the characteristic curves given in Fig. 5.6 apply? Assume 3-in. welded steel-
piping up to the pump and I-in. welded steel-piping to the nozzle, as shown below. 1-in. pipe 2.5 x 10 5 0.0057
3-in.,pipe 8.3 x 10 4 0.0065
Therefore
2
_ (1-)( )(3.56 ) (2)(0.0065)(3.56) 2 L (2)(0.0057)(6)(32.1 2 )
08
EJ - 2 32.2 + (132) (32.2) + U2) (32.2)
= 0.157 + 0.021L + 26.2
Pinch
rolls = 26.4 + 0.021L,
Substituting all available data into the mechanical energy balance, we have
26.4 + 0.021L + 46.2 + 16.0 + L = 104.0,
Sump
or
Solution. From Fig. 5.6, we obtain the total head delivered by the pump, -M*, 15.4
which is 104 ft at 100 gpm. (In reality, the units are ft-lb1 /lbm.) L = l.0 2 = 15.1 ft.
l
Finally, we should examine the net positive suction-head. (We use E J here only olurnes up to several hundred thousand cfm may be involved. Blowers and
between (3) and (2).) (P3 - Pv)/ p ~ P 3 / pin this case. ~urboblowers, on the other hand, are used in situations where larger pressure drops
P3
hnps =- - gL - Ef
p s.2.1 Fans
= 34.0 - 15.l - 0.3 = 18.6 ft. As in the case of pumps, there are many different designs of fans-axial flow,
propeller and cross-flow fans as well as centr~fugal t~pes with a variety of b~a~e
Since hnps > 0, cavitation will not occur with L = 15.l ft. configurations. All manufacturers of ga~-mo~mg eqmpment supply_ characten~tic
5.1.3 Electromagnetic pumps curves describing the performance ofth~ir eqmpment ~nder the specified operatmg
conditions at the inlet, as, for example, m Fig. 5.8, which presents curves for both
In the development of nuclear power reactors, heat transfer from the reactor to
the static and total pressures and the static and total efficiencies. If one wants to
the turbine propulsion system via liquid Na and Na-K alloys has played an
use a fan for moving a specified amount of gas against a given system resistance,
important role. It should be apparent, however, that circulation of the molten
it is necessary to know the static pressure developed by the fan at that flow rate.
~, c: metal is quite difficult to achieve because of the corrosive nature of the metal and
: ~the temperatures involved. In order to overcome these difficulties, an entirely u 14
'different sort of pump has been developed-the electromagnetic pump. ~ 13
ci 12
The de electromagnetic pump operates on the same principle as a de motor. ;; II
The liquid metal, as a conductor, is subjected to a force when a current and a t5 10
90 ~ 9
magnetic field are passed through it at right angles to each other, as shown in 80 ~ 8 ~
&- 70 7 ~
Fig. 5.7. The current connection is made directly to the duct, which is thin-walled, ,.-; 60 6 0
so._that most of the current passes through the molten metal and not the duct wall. g 50 5 fr
JOO ~
The efficiency of such pumps is in the range 10-50 %, in terms of conversion of 2jg ~ 75 ]
applied current and magnetic flux to fluid motion. The flow rate is directly [ii 20 2 50 .l;!
proportional to the current input at the pump duct.
10
0
I fl
oiLL....L..-'-,J20-,-l--3~0,-L-""=--'-+~~-=1::---'--;C~~~,~oo
~
Volume, ft 3 /min X I 0 3
5.2 FANS AND BLOWERS
Fig. 5.8 Characteristic curves for a fan.
Fans and blowers are used for moving gases. We employ fans when the pressure
drop in the system to overcome is not larger than about 50 in. w.c.,* although
First, consider how the fan characteristics are measured. Figure 5.9(a) illustrates
an appropriate system provided with a damper, so that the flow throughput may
be varied. The pitot tube measures the total pressure Pi, which is made up of the
i tQ--:!-~/-
1~1
current-carrier
~Wideopen (b)
Fluid ' - - - - - - - - - Magnet
flow
pole Q
Fig. 5.7 Schematic diagram of an electromagnetic pump for liquid metals.
(a)
* The notation in. w.c., or inches water-column, is often used instead of in. H 2 0, or inches water. Fig. 5.9 Obtaining characteristic curves for a fan.
'i
l
- I
ti
- - - - ----------------------- - - - - - - - - - - - . . ,-----------,:.-011~-AllU-uIUTf"~---~-----
static pressure Ps (measured by the static pressure gage), and the dynamic pressure almost directly proportional to the volume delivered. Characteristic performance
}p i7 2 . By adjusting the damper to various positions, the values of P, and Ps corres- curves for a variable speed turbo blower are given in Fig. 5.10, and a typical blower
ponding to the various flow rates can be generated. Both of these values are relative is depicted in Fig. 5.11.
to the inlet conditions, and are really pressure increases developed by the fan. At
Oil
any flow rate, a theoretical horsepower may be calculated corresponding to either .iii. 12
P or P so that the definition of the static efficiency, r., as well as of the total i
s "
efficiency, r, is based on the actual, or so-called brake horsepower. ~ 10
~
The above statement can also be demonstrated by use of Bernoulli's equation c.
"e;> 8
applied to the system shown in Fig. 5.9(b). Between planes (1) and (3), the energy
'"
.<:
balance is simply -~ 6
0
M* +EI= 0. (5.12) 4
In this case, the system friction is composed of the resistance offered by the 2
damper and the exit losses, that is, the friction between planes (2) and (3). The
mechanical energy balance between planes (2) and (3) is then expressed 8 10 12
cfm X I 0 3 at inlet conditions
(5.13)
Fig. 5.10 Typical characteristic curves for a variable-speed blower.
assuming f3 = 1.
Discharge
Hence, by combining Eqs. (5.12) and (5.13), and noting Pi = P3 , we arrive at
(5.14)
or
(5.15)
Writing the above expression with the actual or brake horsepower P8 and the
fan's efficiency r, we obtain
rP8 = -M*pQ =
[{Pz - P1) + -pVz2]
- Q. (5.16) Impeller Inlet
2
Fig. 5.11 A Roots pump or turboblower.
Thus the total power delivered is based on the sum of the static pressure
(P2 - Pi) and the dynamic pressure (}p i72 ), or what we call the total pressure., as
5.2.3 Interactions between fans and systems
incorporated in brackets.
Since a fan is part of an overall system, the system as a whole determines the size
of the fan required. For any system, there is a certain curve of volume flow versus
5.2.2 Blowers
system resistance or pressure drop (see Fig. 5.12). The reader should realize by
In order to produce large heads to overcome large pressure drops in systems, either now that the resistance usually increases as the square of the volume flow. The
i~ positive displacement or centrifugal blowers are used. Blowers and turboblowers system may contain any combination of duct .work, beds of fluidized or packed
:( find their typical metallurgical applications in producing the air blast for bl~st solids, dust collectors, flues, etc. There is usually a specific volume throughput Q',
furnaces and cupolas. required for proper operation of the process; this fixes the pressure drop resistance
Blowers are essentially constant-pressure machines with power consumption . of the system, which in turn must be overcome by the fan or blower. Essentially
Table 5.1 Variables for fan laws
then, the fan characteristic curve must intersect the system curve at the desired
Q'-11P coordinate (the so-called operating point). The efficiency and required Speed of rotation n
horsepower are then fixed. Note that a different fan with a different characteristic Impeller diameter D
curve would place the operating point at a different position on the system curve, Gas density p
resulting in a flow different from Q'. Static pressure PS
The normal operating range for a fan is to the right of the peak of its pressure- Power PB
volume curve. The so-called pumping limit, defines the furthest possible left-hand Volume flow Q
operating point. Pumping occurs when the fan is slowed down to the point where
the static pressure created is less than the static pressure in the discharge line. At The fan laws are:
that point a flow reversal takes place. After this reversal, a momentary drop in the (5.17)
I,'
!1,, pressure in the discharge line occurs, and flow starts in a normal direction again.
" '~1,
This pattern is repeated very rapidly and results in a pumping-type action, which Ps = k 2 D2 n 2 p, (5.18)
1l
~1: may cause damage to the fan unless flow below this limit is avoided. It usually
turns out that this point is reached at the maximum on the fan total pressure curve,
which in turn is usually slightly to the right of the maximum of the static pressure
J1 = k3 D 5 n 3 p.
Here is an example of the use of the fan laws. If the density of the gas varies, and
(5.19)
,,! the speed and volume flow remain constant, then Eqs. (5.18) and (5.19) tell ~s that
curve.
,, It is not always possible to make desired changes in the operating point of a both the horsepower and the static pressure would be expected to vary directly
,[ I
,j
I' fan system. Changes in operating temperature or pressure drop require reevalua- with the density. The proportionality constants k 1, k 2 , and k 3 are constants over
a limited range, which should not deviate too far from a given point on the curve
I tion of the system resistance curve, and if this changes, the operating point shifts,
l [,!
I'' '
unless the fan is adjusted to keep Q' constant. Conversely, if it is desired to change
Q', then the operating point has to be changed, and consequently the fan character-
istic curve must be changed.
of pressure versus flow rate. . .
Changes in the pattern of flow of the gases entermg the fan will also change the
characteristic curve, since the calibration is performed with gases entering smoothly
at right angles to the rotor. If the gases enter otherwise, then their flow through
the fan will affect the characteristic curves.
;i
'I I
ii Ei I,, &
Pumping limit
~It
~(
&j' ! IO
t:; I fJ
I ~ JOO
i I ~
. ~ 90
'
::1 "'----Operating pressure
Operating
~ 80
'1 "i:l efficiency
~
"'
c::
70
t 60
50
Operating 40L---1---1--1---1...-L::...lLL-----1.-.L.-.L..L-LL--'-'L....-L~
horsepower
"" 120
:;) JOO
~
8, 80
Volume flow rate, Q 1;\
::; 60
..c::
Fig. 5.12 Relationship between the fan curves and system curve. .,,, 40
"
~ 20
In order to estimate the effect of deviations from the specified conditions, or o~~~~~~--7';--,,~~-o'nc--;:;>;;--~~;;-;-:~
O IO 20 30 40 50 60 70 80 90 JOO !IO 120
the effect of a change in fan size or speed, on the behavior of the fan or the gas, the Rated volume at constant density,%
so-called fan laws are used to convert from one set of operating variables to Fig. 5.13 Effect of dampers on the characteristic curves of a fan.
another. The variables are given in Table 5.1.
L
------------------------------.
hen added to the calculated design bed-resistance of 9.5 in. w.c. at the design
flow, gives a total system design resistance of 13.0 in. w.c. The fan to be used in
this system is then specified to meet a 320,000 cfm flow at 250F with a 13.0 in. w.c.
static pressure drop (point 0 in Fig. 5.15).
Suppose, however, that when the plant is built, the actual pressure drop across
the bed of pellets at the required flow is 21.0 in. w.c. This means that the actual
system curve (I) is different from the design system curve. Now the fan operates
at point 2 (Fig. 5.15), and it pulls less than the required flow through the system
(280,000 cfm). In this case, the only alternatives to regain a flow of 320,000 cfm
- are either to decrease the system resistance, or to change the characteristic curve
by increasing the fan speed. In the first instance, one could decrease the system
Fig. 5.14 Schematic diagram of an iron-ore pelletizing plant.
resistance by decreasing the bed height. If this is done, however, the lateral speed
of the bed must be increased in order to keep the production rate (lb solids treated/hr)
We can accomplish control of the flow by varying the speed, by using inlet
at the design value. However, this may not be possible, since there is usually a
guide vanes, or by placing a damper in the system. The effect of guide vanes, or
minimum reaction-time for the exposure of the solids to the hot gases, and the
the placing of a damper in the system, on the rated (without vanes) capacity is
shown in Fig. 5.13. original design is usually close to that minimum. Changes in duct work may also
be possible, but usually are not practical once the plant is built.
It is best, therefore, to overrate fans somewhat on both the pressure and volume
The other alternative-speeding up the fan until the operating condition is at
specifications, but not too much, since severe dampering, if necessary, decreases
point 3-is easier but may be very expensive. Reference to the fan laws shows
the efficiency and wastes the available horsepower. The problems that can arise
us that the power required is proportional to the cube of the flow; that is, for
from errors in matching the system and fan curves are illustrated in the following
example. constant fan-size and gas-density, Q oc n, and PB oc n 3 , so that PB oc Q3 . Therefore,
for an increase of flow in the ratio 320,000/280,000 = 1.14, the ratio of the new
Example 5.2 Part of an iron-ore pelletizing plant consists of a moving chain grate horsepower required to the old will be (1.14) 3 = 1.50; in other words, a motor 50 %
(carrying a bed of pellets subjected to hot gases) and associated gas cleaning and more powerful is required. This may be quite expensive and could cause many
exhaust equipment, as shown schematically in Fig. 5.14. The total pressure drop problems in the electrical system. The ultimate so!Ution is to buy a new fan.
due to duct work, at the required flow of 320,000 cfm at 250F, is 3.5 in. w.c., and, If, on th.e other hand, the actual bed-resistance were, e.g., 7 in. w.c., then the
fan would pull too much gas through the system, as at point 1 (Fig. 5.15). This
can easily be corrected by increasing the system resistance through the use of
dampers, unti~ the system and fan curves coincide at the required flow and pressure
drop, point 0.
L
~". igb-ve\ocit)' iets
159
"'-.,
flow of an ideal compressible gas under these conditions yields an equation for the The Mach number, M, is defined as V/V,, so that finally
velocity in the nozzle opening Pi:
dV dA
v (M 2
- 1) =A. (5.27)
(5.20)
Now, if the velocity at any point in the nozzle is less than M = 1, and if the area
where reference points are defined in Fig. 5.16, and y = Cp/Cv. We may use this of the nozzle decreases at that point (dA/A = negative), then the velocity increases
equation for a flow nozzle operating at subsonic velocities. at that point (dV/V = positive).t At the throat, (dA = 0), either M = 1 or dV/V =
:A t
Po
l~, P,
V=O
~J,Po,f>o C:.P 1 t.r.
I
I
I
i
_, I
Pt ---------j--
,,......
I
____
I
I
I
I
I
v = VfP\aP)
s \EJP} ~onstant (5.21)
1
I
I
(b)
entropy I
I
For an ideal gas,
I
i; I
v=
s vfYP.
---;; (5.22)
.0
E
::;
-"
I
I
I
________ J.I __
V"
e
'-' (c)
The equation of momentum for one-dimensional flow is ::>'". I
I
.':' I
dP - _ ~ I V"e
-+ VdV= 0. (5.23) 0
-;:; I
v,
p > I
0
2
Substituting dP = V, dp into Eq. (5.23), we get
~
~
vs2 dp
p
+ vdV = o. (5.24) ~
~
"
20
(d)
The continuity equation, which requires that the mass flow rate remain constant
at all points in the nozzle,
dp + d}? + dA = O (5.25)
Pt
p~
P/ P0
p V A '
Fig. 5.17 Reference points and schematic internal conditions for various operating pressure
must be satisfied along with Eq. (5.24), and so ratios for converging-diverging nozzles. ... ~
d}?(V
V V,2
2
- l) =dA_
A
(5.26) t The reader should consider whether or not M > 1 can be achieved in the converging portion
of a nozzle.
- - - - - - - - - - - - - - - - - -------------------------------;:,= ---------rngn:;ve1ocurJe~101
0. I~ M < 1 at the throat, then df7/f7 is zero and no further increase in velocity is portion of the nozzle should be made in order to produce. a jet with exit press~re
possible, i.e., this is the maximum velocity which can be achieved. p , equal to ambient pressure. Such a condition establishes the most effective
If M = I at the throat, then we have reached the point at which supersonic c~ndition, as discussed in Section 5.3.2.
flow can take place. By adding a diverging section on the end, and referring to Up to this point, it has not been established wh~t the ab~olute ~ressure at ~he
Eq. (5.2:), we see that if_dA/A is positive, then either dV/V will be positive, and a throat, relative to the exist pressure, is. Several possible solut10ns anse, dependmg
further mcrease to higher velocities is achieved in the diverging section, or the on the nozzle design. There are two values of Pt* IPe-and therefo~e two ~alue~ of
fl?wm.ustcomet~astop. T~efo:mercaseoccursaslongasPt > Pexit Aconverging- Pe, shown in Fig. 5.17(b), which result in isentropic shockless flow m the dive~gmg
d1vergmg nozzle 1s shown m Fig. 5.17. It is usually called a deLaval nozzle. ortion of the nozzle for the conditions where M = 1 at the throat. The higher
For a converging nozzle, we obtain the conditions required to attain sonic ~ressure P; will cause the fl.ow to b~come subsonic again im~e~iately at the thro~t,
velocity at the throat by using the energy relation for an ideal gas undergoing and will cause a decrea~e m velocity; the lower pressure Pe will allow supersomc
adiabatic flow:
flow throughout the nozzle. If the exit pressure is between P; and P;, and Pt = P[,
T.o then somewhere in the nozzle, a discontinuous transition from a lower to higher
:j ~
I' y-1
- = 1 +--Mz (5.28a) pressure (and simultaneously from supersonic to subsonic flow) occurs, as shown
1
T 2
schematically in Fig. 5.17. This produces a shock wave.
1;'I.I
!''
~hen !"1 = 1, sonic conditions exist, denoted by an asterisk, and since the flow If the nozzle design is such that Pe = P;, then we may proceed to calculate the
is also 1sentropic, substitution of the ideal gas relationship exit velocity, using the equation
!__ = (!__) y/y-1 = (!_).le (5.32)
(5.28b)
Po Ta Po
gives the critical pressure ratio
or
~: = (y ~ lry- l. (5.29)
2
M 2 =--
e Y- 1
[(p
-
Pe
0 )<y-l)/r J
-1 (5.33)
This means that the ratio of reservoir pressure to throat pressure at the sonic flow
The mass flow rate and throat area are still given by Eq. (5.31).
condition is governed only by the value of y. For air and oxygen, Pi*/ P0 = 0.528.
Now that we have the mass flow rate, we may calculate the area of the exit by
F~~ pressure ratios between 1.00 and 0.528, the mass flow rate of the gas is W, =
P1J/;At, and using Eq. (5.29b), means of a mass balance, since W,* = it;,,
'. _ 0 0- 1 -
W, - At - -
2p P
y - 1
y[ (pt)(y-l)/y](pt)l/y
-
P0
-
P0
. (5.30) At PoPoY ( y
2
+
)(y+ 1)/(y-1) _
1 - Ae
J ( - 2) [ _ (Pe) (y
PoPoY y - 1 1 Po
l)/y] . (Pe) l/y
Po ,
Any further decrease in the ratio Pt! P0 below that for Pi*/ P0 , caused, for example, (5.34)
by increasing the reservoir pressure P0 , will not cause a further increase in the mass
flow rate. out of a converging nozzle, since the condition of isentropic flow is no (5.35)
longer satisfied beyond the nozzle. If (Pt! P0 ) :::;; (E'r* / P0 ), then the nozzle is said to
be choked, and the mass flow rate is
Ideally ' we want Pe = Pam b.ient' since then the ultimate adiabatic expansion is
2 )(y+l)/(y-1) reached, and the jet issues from the nozzle at atmospheric pressure. If Pe is less
W,* =At PoPoY ( - - . (5.31) than Pa, the ambient atmosphere compresses the flowing jet and collapses it in a
y+ 1
series of shock waves. If Pe > Pa, the jet continues to expand beyond the nozzle
For the nozzle with the additional diverging section, the mass flow at sonic or tip. In both non-ideal situations, the efficiency of conversion of the nozzle velocity
supersonic conditions is still given by Eq. (5.31). However, since the velocity to jet momentum decreases. .
increases along the length of the diverging portion, the pressure correspondingly The angle of divergence of the nozzle is usuhily about 7 in order to avmd
decreases from that at the throat. The problem is to determine how long this separation of flow from the nozzle walls.
L
-~------.
Example 5.3 In the basic oxygen steelmaking process, we decarburize a bath of 5.3.2 Jet behavior
molten iron-carbon alloy with gaseous oxygen which is blown into the bath from
As a jet exits from the nozzle, it entrains adjacent slow-moving air, ':"hich in t1:1rn
a lance held above the bath. Determine the dimensions of the converging-
acts as a drag, creating turbulence. This slows some of the supersomcally flowmg
diverging nozzle which are required to achieve a velocity of Mach 2 at the exit
gas to sonic and subsonic velocities and the supersonic ~ore of the jet gradua~ly
with an oxygen flow rate of 15,000 scfm.* What is the required driving pressure?
decays, until at some distance from the nozzle the core disappears, and the enttre
Solution. From Eq. (5.33), we can obtain P0 by assuming that Pe = 14.7 psia to jet is subsonic. Figure 5.18, ?ased o? th~ work of Anderson and Johns, 1 shows the
achieve the best behavior of the jet, and noting that y for 0 2 is 1.4: relationship between the exit velocity (m terms of Mach number) and the length
Po) o.2s6 _ 22 of the supersonic core (in terms of either the ratio x/dt> where x is the distance
(-Pe ---+
(...1....)
1.0, from the nozzle and dt is the throat diameter, or the ratio x/de, where de is the exit
0.4
diameter).
.. t and thus
Po
- = 7.85. >-< 1.,, 50
Pe i:
0
u
u 40
For Pe= 14.7 psia, P0 = 115.3 psia. Now we can make use of Eq. (5.35) to find a
~
the ratio At/Ae: lil
0.
i;l 30 30
At)2 = (2-)(2.4)2.4/o.4(~)2;1.4[l _ (~)o.4/t.4] '-
0 x
( Ae 0.4 2 115.3 115.3 ..<::: d,
Oii
20
""'
~:)
~
( = 0.595.
Finally, we can calculate At from Eq. (5.31), since the nozzle is choked, as otherwise
supersonic velocities could not be reached. OL__ _J___ _L __ _ . _ _ _. L __ _ . __ _.J____ _,O
The mass flow is 1.0 1.4 1.8 2.2 2.6 3.0 3.4 3.8
Mach number
lbm 1 min
W* = 15,000 scfm x 0.089 - f3 x - Fig. 5.18 Relationship between the length of supersonic core and exit Mach number. (Adapted
t 60 sec
from Anderson and Johns by Smith and by Holden and Hogg, ibid.)
= 22.2 Jbm/sec.
2 6
22 _2 = At (0.650 lbm) ( 11~.3 lbJ 144 in ) ( l.4) (2-) (32.17 lbm-ft) We measure the overall spreading of the jet either by the ratio of the velocity
3 2 2 lbrsec 2
ft m ft 2.4 V,. at~ given radius r (measured from the center line of the jet) to the center-line
4 velocity V,,, or by the ratio r0 /rt> where we define r0 as the radius at which the
= AtJl6.25 x 10 ,
velocity is one-half that at the center line. The typical velocity or impact pressure
At = 0.0551 ft 2 profile is that of a normal distribution about the center line. However, the spreading
= 7.94 in 2. of the jet is minor until the supersonic core has decayed. Once this point has been
reached, the jet expands at an included angle of about 18. Figure 5.19 indicates
Thus the th:\"oat diameter dt = 3.19 in., and the exit diameter de = 4.14 in. the spreading profile of two jets, and Fig. 5.20 gives a dimensionless graph for
determining the width of the jet as a function of the exit Mach number and the
---~ __r-=4.14in. distance from the nozzle. W.e then take the effective jet radius r as 2r 0 , and can
Po
calculate it at any distance from the nozzle with the help of Fig. 5.20.
+,?t Because of the entrainment of more gas into tne-jet, the mass of the jet increases
as we get further from the nozzle, but the velocity decreases. For supersonic jets,
L
* Standard cubic feet per minute. 1
A. R. Anderson and F. R. Johns, Jet Propulsion 25, 13 (1955).
rngn-vemcrry)l'IS--ro;,
e impact pressure at the center line (virtually entirely dynamic pressure) has been
0 2
rrelated as a function of the distance from the nozzle; Fig. 5.21 shows this
IO I I rrelation. For subsonic jets, the velocity at the center line has been observed as
I I Supersonic ;~oportional to (d,/x). 3 Due to the f~ct that the impact pressure is essentially
20 I I core
I I proportional to the square of the velocity, we would expect
I I
k(~)-
30 2
Mach I I Mach . pc = (5.36)
I \,I
40
I
I V'
\
Po - pa d,
in. 50 I I When we test this against the supersonic data in Fig. 5.21, we find that at Mach
I \ . number 1.5, the relationship still holds, but that for higher Mach numbers, the
60 I I
I \ slope is no longer - 2.
70 I I
I
I \ o..."
I
J.0
80 \ 0.8
I ~
0
90 I
I \ ~
0.6
\
JOO I -'-"----'
0.4
20 15 IO 5 0 5 10 15 20
in.
0.2
Fig. 5.19 Spreading of Mach 1 and Mach 2 jets with identical flow rates of 6500 scfm 0 2 .
(From Smith, ibid.)
0.10
0.08
0.06
0.04
0.02
o.oi
I 2 4 6 8 10 20 40 60 80 JOO
x/d,
Fig. 5.21 Maximum impact pressure as a function of Mach number and distance from the
nozzle. (Adapted from Anderson and Johns by Smith, ibid.)
The increase in the mass fl.ow of a subsonic jet due to the entrainment of the
surrounding gas has been found to be directly proportional to the distance from
the nozzle exit, according to the equation
(5.37)
20 40 60 80 JOO
x/d, 2
G. C. Smith, J. Metals, 846 (July, 1966).
3W. G. Davenport, D. H. Wakelin, and A. V. Bradshaw, article in Heat and Mass Transfer in
Fig. 5.20 Jet-spreading characteristics as a function of.Mach number and distance from the
nozzle. (Adapted from Anderson and Johns by Smith, ibid.) Process Met., Inst. of Mining & Met., London, 1967.
i
l-
~--- ,, -
where Me = jet momentum at nozzle exit ( = iv., V.,), x/d = distance from the from Fig. 5.20, at Mach 2.0, x/d1 = 31; x = (31)(3.19) = 99 in. above the bath.
nozzle exit, iv., = jet mass flow rate at nozzle exit, Wx = jet mass flow at x, Pa = The impact pressure is found from Fig. 5.21:
density of the ambient gas.
In supersonic jets, the entrainment in the region where supersonic flow - -pc- = 0 . 09,
Po - Pa
predomin~tes is less than in the subsonic region, where Eq. (5.37) is satisfactory.
A :ery sa~1~factory representation of the increased jet mass for supersonic jets with P0 - Pa = 115.3 - 14.7 = 100.6psi.
exit veloc1t1es between Mach 1 and 2 is given by*
Pc= 9.06 psig.
Thesupersoniccoreisdissipatedatadistancex = 13(de) = (13)(4.14in.) = 53.8in.
from the nozzle.
Supersonic jets in metallurgy have thus far been applied mainly in the basic
0
oxygen steelmaking process, as mentioned in the examples above. The inherent
where T0 = stagnation temperature, 0 R, 7;, = ambient temperature, R, and rate of the reaction is so fast that the rate of decarburization is limited only by the
(x/de)core = length of the supersonic core determined from Fig. 5.18. This equation rate at which oxygen is supplied to the bath of pig iron. The rate of supply is a
does not describe the increasing entrainment in the region from the nozzle to a maximum in the case of the converging-diverging nozzle operated at supersonic
distance of about 10 nozzle diameters downstream, but since this is not the region conditions. The jet of oxygen penetrates into the liquid, displacing metal. The
of usual interest and also the rate of increase is not very strong in this region, the distance to which the jet penetrates has been studied,4 5 but the general correlation
equation is still useful. Note that as the ambient temperature increases, the
is not well enough established to warrant its inclusion at this time.
entrainment decreases. Subsonic jets of gas are often injected into molten metals, such as in the older
We should point out that the entrainment of the surrounding gases dilutes practice of oxygen lancing of steel, but this is usually done with a straight pipe, and
the jet gas-concentration. For example, the concentration of oxygen in cold jets the analysis above is not applicable here. For a discussion of the behavior of such
of pure oxygen issuing into cold air, at a distance of 6 ft from the nozzle, is 90 %in a jet, the reader can consult the review articles by Holden and Hogg and by
6
the case of.a Mach 3 nozzle versus 80% in the case of a subsonic nozzle with 5
Davenport et al.
25 psig driving pressure.
Example 5.4 At what height above the bath should an oxygen lance discharging 5.4 VACUUM PRODUCTION
15,000 scfm 0 2 at Mach 2 be placed in order to have an impact area of 400 in 2 In recent years many metallurgical processes have been developed in which vacuum
on a bath of molten iron-carbon alloy? What is the impact pressure at the center plays an important role. These include vacuum annealing, vacuum deposition of
line? How long is the supersonic core? coatings, vacuu!Jl melting, and vacuum degassing. In addition, vacuum equipment
2 is a standard item in almost every metallurgical laboratory. For these reasons, it is
Solution. The effective radius of the jet in order to cover 400 in is
import~nt that the metallurgist be acquainted with the principles and operation
r = j400/3.14 = 12.72 in. of vacuum-producing equipment.
The most important variable in the design and specification of equipment for a
Since r = 2r 0 , vacuum system is the pressure which the pumping system must be able to maintain
r0 = 6.37 in. in the work chamber. In vacuum technology the standard unit of pressure is the
torr.* It is defined as the pressure exerted by a column of mercury 1 mm high and
rt = 1.595 in. from Example 5.3.
Thus 4
R. A. Flinn, R. D. Pehlke, D.R. Glass, and P. 0. Hays, Trans. A.l.M.E. 239, 1776 (1967).
5
r0 /r 1 = 6.37 /1.595 = 3.99, W. G. Davenport et al., ibid., pages 5-22.
6
C. Holden and A. Hogg, J. Iron and Steel Inst. 198, 318..(1960).
*The name torr comes from E. Torricell~ a student of Galileo, who devised the first single
*This equation is derived from data presented in the report by J. D. Kapner and Kun Li, stroke pump by inverting a closed tube of mercury into a dish containing mercury, creating a
Mixing Phenomena of Turbulent Supersonic Jets, American Iron and Steel Inst., June 26, 1967. vacuum in the tube.
L
----------- ----- ---------- ------- - - - - - - - - - - - - - - - - - - - - - - - - - - - . .
measured at 0C, sea level and a latitude of 45. A vacuum of l torr is no longer
considered to be a particularly good, or hard vacuum, since now we are able to
Another variable is the speed at which the desired pressure is reached. We
define the speed SP, of any type of vacuum pump by
!
produce much lower vacuums. Figure 5.22 compares the various pressure scales,
I indicating the ranges of application of various types of vacuum gages and pumps. SP= Q/P,
where P is the pressure at the inlet to the pump, and Q is the throughput at that
(5.39)
Vacuum domain
point. Typical units for SP are liters/sec, for P, torr, and for Q, torr-liters/sec. We
\-- Very high
I0- 9 10-s 10-1
--J- High
10-6 10-s 10-4
Medium - + L o w
I0- 3 10-2 10- 1 1 IO I02 10 3
-J may use the same definition to express the pump-down speed of a system, Ss,
consisting of a working chamber, connecting duct work, and pumps, but the speed
Pressure in torr (mm Hg) depends on the design of the rest of the system as well as on the pump itself.
I I I I A system, initially at atmospheric pressure, is first roughed out by either
10-6 10-s 10- 10-3 10-2 10-1 I0 1 102 I0 3 I0 4 I0 5
mechanical pumps or the first stage of an ejector system until a pressure is reached
rrT"r'T"'T"r
viscosity is no longer important in determining the flow rate, and the gas flow
becomes molecular flow. At an intermediate pressure, a transition flow regime is
encountered when the mean free path is of the same order of magnitude as the
107 10 8 I0 9 10 10 10 11 Ion 10 13 10 14 10 15 10 16 10 17 10 18 I0 19 equipment dimensions.
Density in molecules per cubic centimeter (at 25C) 5.4.1 Molecular flow mechanics
I I I I I
Operating pressure range In the molecular flow regime, the gas molecules move randomly, with a Maxwell-
for vacuum pumps Boltzmann distribution of velocities, and collisions between mo~ecules are rare
compared with collisions with the walls. The only transfer of momentum is
between molecules. and the wall instead of from molecule to molecule. Net flow
results from the. statistical effect that the number of molecules leaving a given region
Vapor diffusion pumps is proportional to the number of molecules in the region, and the number reentering
Operating pressure range will be proportional to the number in the adjacent region. Knudsen developed the
of vacuum gages
Fig. 5.22 Comparative pressure scales and ranges of application of various vacuum pumps Fig. 5.23 Schematic chamber being evacuated from an internal pressure of P 1 to an external
and gages. pressure of P2 through an aperture of area A.
. -c----- ------- ---------vacnum-prooncnon--m
equations which govern the flow of gases through various geometries under Table 5.2 Conductances of various geometric shapes for molecular flow*
molecular flow conditions. Consider a chamber with an aperture of area A, as in C, for air at 25C
Sha pet Conductance (liters/sec)
Fig. 5.23. The pressure on the downstream side of the aperture is P2 and that in
the chamber is P 1 . If the concentration of gas molecules is given by n (molecules/
cm 3 ), and the number hitting the wall is r;;l
~ C=3.64A
(y)112
M = ll.7A
Z = nV [molecules), (5.40)
4 \ cm 2 -sec
where Vis the average molecular speed given in Eq. (1.4), then the frequency z Az(y)112 D3
of gas molecules per unit area passing from the high pressure side of the aperture C=19.4-- = 12.2-
BL M L
\\ ,~] through it, is
Z1 -
-
~(2KsT) 1/2
C.. - '
molecules
(5.41)
2vn m cm 2 -sec
The frequency from the low-pressure side back through the opening c= D3(
3.81- -
T) 112
L M
- ~(2KsT)
Z2 - C.. -
1/2
' (5.42)
2vn m
yields a net frequency of c= 9.70
b2c2 (
-
T) 112 =
b2c2
31.1---
(b + c)L M (b + c)L
Znet = Z1 - Z2 =
1 (2K--;;;-T)
Jn 8
1 2
1
(n1 - n1). (5.43)
2
D2
For a circular. opening, the net rate at which molecules leave the chamber is y)112( 1 ) 9 4
C = 2.85D2 (M 1+ 3L/4D = .1 1 + 3L/4D
nD
R = 4
2
Jn(2K T)
Znet = - - ---;---
1
/2 2
D (n 1 - n 2 ). (5.44)
8
_\__
- - - - - - - - - - - . . - , . . . . ,... u:u:1-p~u'--.:1u11---1-r;,-----
Table 5.2 presents the conductances of other shapes besides apertures. the system is isolated, compressed, and discharged to the atmosphere with each
Pumping speed has the same units as conductance, but it should be noted that rotation of the piston. These pumps have intrinsic speeds, S0 , ranging in value
conductance implies a pressure gradient across a specific geometry. Pumping from 0.5 to 350 liters/sec.
j.J'
'
speed is simply the volume of gas flowing across any plane in a system per second, From vacuum system
which is measured at the pressure existing at that particular plane.
Discharge
The pump-down speed of a system depends on both the pump speed and the valve
conductance of the connections. Refer to the schematic diagram of a laboratory
vacuum melting system in Fig. 5.24. Since P = Q/S at the inlet to the duct,
PP = Q/Sp at the inlet to the pump, and (P - Pp) = Q/C over the duct length, then Oil
1 1 1
(5.49)
or
s-s(
- P
1
1 + SP/C )
(5.50)
This means that the effective pump speed S of a system being evacuated by a pump Fig. 5.25 A mechanical pump of the rotary, oil-sealed, vane type.
with rated speed SP, cannot exceed SP or C, whichever is the smaller. If the con-
ductance of the duct is the same as the pump speed, then S = SP/2. This should . There is a lower limit to the pressure that a pump may produce, known as the
emphasize why it is desirable to make connections between the working chamber ultimate pressure Pa, at which point the speed drops to zero. This limit is deter-
and the pump as short and as wide as possible. mined ?Ythe amount of back leakage of a very small quantity of gas Qa, which is
nearly mdependent of pressure. At the inlet to the pump,
Vacuum induction
melting chamber
SP = Q - Qa = So ( 1 - Qa) (5.51)
I pp Q
c Roughing valve At the ultimate pressure, Qa = Q and S0 = Qal Pa. Therefore
S,P Backing \
~:)
valve
SP = s0 ( I - (5.52)
if ,1 Valve
where
5.4.2 Mechanical pumps
Most mechanical vacuum pumps are of the positive-displacement, rotary-piston S' -
-
s0 (1 + IS )
/C .
type with sliding vanes, and sealed with oil (Fig. 5.25). A small quantity of gas from 0
'
L
------------ ------ .. ------------------r---"'"""'V'---~-----
Single-stage pumps have ultimate pressures in the range from 10- 2 -10- 3 torr. Pump inlet (low pressure)
Two-stage mechanical pumps can reduce the ultimate pressure to 10- 4 -1o- 5 torr.
In order to reduce the pressure from any water vapor in the system, a trap, either
cold or chemical, should be placed ahead of the pump to remove moisture from the
incoming gas stream. ~;i1~r cooling { High forepressure
to mechanical forepump
If a slightly lower pressure is desired, and if the quantity of gas is large, a Roots First
pump (shown in Fig. 5.11) may be used in conjunction with an oil-sealed mechanical compression
stage
forepump. This combination may produce an ultimate pressure of 10- 6 torr.
Typical characteristic curves for various types of mechanical pumps are Baffles to
trap pump
shown in Fig. 5.26. Second {
fluid
compression
stage
7 100 Third
u
~ compression {
stage
iI
40
~ Vapor condenses
;;_
20
~;~~~:~ID{
"'"'
"O
10
"'0. I
using the previously defined nomenclature. Then, the intrinsic pump speed S0 is where C1 is the conductance of the trap. For well-designed traps, SP ~ S0 /2. The
So=HZA=g_=HAV, ultimate pressures attainable with well-trapped mercury diffusion pumps are of
(5.56) the order of 10- 6 -10- 7 torr. (A pressure of 10- 13 torr is reported to have been
n PP 4
developed with a liquid nitrogen-trapped, three-stage mercury pump constructed
and, substituting Eq. (1.4) for V, we obtain of glass.) More commonly, low-vapor-pressure hydrocarbon oils or silicone oils
are used as the working fluids. They are capable of producing vacuums of 10- 6 torr
S0
- _!!__(2KBT)
- C
1/2
A, (5.57) without traps, since their inherent vapor pressures at room temperature are of the
2yn m
order of 10- 10 torr.
or
5.4.4 Pumpdown time
112
S0 = 3.64H(:) A, liters/sec. (5.58) Returning to our laboratory vacuum melting unit, the basic equation relating the
change in pressure in the tank to the pumping speed of the system, is
Specifically, for air at 20C,
dP
PS= - V dt + Q1 , (5.61)
S0 = l 1.6(HA), liters/sec, (5.59)
where A is measured in cm 2 The coefficient H is a measure of the collection where P = pressure measured at a specified point in the system, S = speed at that
efficiency of the pump and is called the Ho coefficient (named after T. L. Ho). For point, V = system volume, and Q1 = additional gas flow made up of the leak rate,
most diffusion pumps the Ho coefficient is about 0.5. interior surface outgassing, and any process gases. In well-maintained systems
These equations imply that S0 is independent of the pressure for diffusion with no process gas evolution, Q1 is eventually brought to a negligible level.
pumps, and this is nearly true, as shown in Fig. 5.28. Solving Eq. (5.61) for Sand equating with Eq. (5.54), we obtain
The ultimate pressures attainable with diffusion pumps depend on their
dP S'
design, including the number of stages, and also on the working fluid used. Ifwe ---= --dt. (5.62)
use mercury, a vapor trap, in the form of a baffle system externally cooled to low (P - Pa) V
temperatures, must be placed between the diffusion pump and the system being We find the pumpdown time by integrating this equation from the initial tank
evacuated in order to prevent back-diffusion of mercury vapor (Pttg = 1 x 10- 3 torr pressure P1 to the final pressure P2
at 20C). In this case, the effective speed of the pump is
2.30 V log (P1 - Pa) (5.63)
(i + ~o/CJ
t= --
(5.60) S' P2 - Pa
SP= s0
I
I
I
diffusion pump. Figure 5.28 includes the performance curves for a diffusion pump
and a mechanical pump. In operation, the forepump is turned on, and run alone
, until a forepump inlet pressure less than the limiting forepressure of the diffusion
I pump is reached, at which point the lower stages of the diffusion pump become
P0 10- p px 10 operative. If the pressure at the diffusion pump ifllet is P, then the throughput is at
Pressure, torr point Q, and since the throughput is the same at any instant for both pumps, then
Fig. 5.28 Characteristic curve for a typical vacuum diffusion-pump, with a matching me- the forepump inlet pressure is PX, its speed is SX, and its throughput Qx.
chanical forepump's curves included.
l,,c s;nce the speed and thmugh put requ;red of mechankal pumps when employed
-~--~------- - --------
Problems 179
as forepumps are often rather low values, their use in large systems for roughing out The pumping capacity or throughput of a steam ejector is generally given in
would require excessively long pumpdown times. Therefore separate mechanical terms of pounds of dry air removed per hour. For comparison purposes
roughing pumps are often used, and then turned off when the diffusion pump-
forepump combination can be applied. I lb air at 68F /hr =79.5 torr-liters/sec.
Figure 5.30 illustrates the range of pressures and throughputs obtainable with
First typical combinations of ejector stages and condensers.
Suction chamber stage Diffuser
! Nozzle ~ 300
l E 200
E_ 150
"'
"
:2 100
IS. 80
Inlet port, Pe ::;"' 60
"
~ 30
AB~
20 - A JiB
Second stage
ABC~
Velocity
A B
c
A-X-C flIDQ C
~~
I ABX CCJ1"L1:C
0.6
I
~x~c
~---Water discharge
I
Fig. 5.29 Three-stage steam ejector with intercondenser. Schematic diagram of pressure-
velocity relationships in the first-stage ejector is included.
A-X-B-X-C
Assume same steam consumption
100-psig steam and 85F water '
:::~
B B C
AB-X-B-X-C
5.4.5 Ejectors 0.06
ABB-X-C-X-C X X
0.04
For very large systems, other types of vapor pumps, called steam ejectors, are used. 0.03
Figure 5.29 shows a schematic diagram of such an ejector. The steam is made to c
0.02
pass through a converging-diverging nozzle, designed to reach Mach 2 or higher
. I' at the nozzle exit and, correspondingly, a very low pressure Pe. The inlet port
O.Ql5
0.01._____._ _J _ _ _ . 1 __
ABB-X-C-X-C
__,___ __,___ _ , __ _ , __ _ _ ,_ ___._ _J _ __ _L___J
design pressure is then this Pe, and the steam jet entrains gas molecules at this 0 10 20 30 40 50 60 70 80 90 100 110 120
pressure, as it leaves the nozzle. Flow into the port is again due to statistical Ory air load, lb/hr
molecular flow, at the lower pressures. Once entrained, the gas and steam slow Fig. 5.30 Ranges of application of various steam-ejector (A, B, C)-intercondenser (X)
down and are compressed in the diffuser so that they may exit at an exhaust pressure combinations. (From an article by F. Berkeley, Chem. Eng., April, 1957, page 255.)
equal to the atmosphere, in the case of a single-stage pump, or at the design inlet
pressure to the next stage in the case of a multiple-stage ejector. In order to lighten PROBLEMS
the load on the succeeding stages, condensers are often inserted between stages to 5.1 Design the system and specify the pumps which are needed to evacuate a laboratory
remove the steam from the preceding stage.
l
melting chamber of 30-ft3 volume to 10 pressure in 20 min. Discuss how your design would
180- Flow and-vacuum production
l. 181
~-------- ---- -- -------------
The nature of heat transfer by radiation is quite different from that by conduc- / /
tion, and consequently the basic rate equation for heat transfer by radiation is
in no way similar to Fourier's law. In Chapter 11, we shall analyze radiation
JIi problems in detail; here, we state briefly that thermal radiation is part of the 6
electromagnetic spectrum. The energy flux emitted by an ideal radiator is propor-
tional to the fourth power of its absolute temperature FOURIER'S LAW AND THERMAL
eb = aT 4
,
CONDUCTIVITY OF MATERIALS
1
i'
I!
where eb is the emissive power (a special term for thermal energy transferred by
radiation) and a is the Stefan-Boltzmann constant. The processes of conduction
i .and radiation frequently occur simultaneously, even within certain media. In many
practical situations, however, one mode is negligible with respect to the other, and Thermal conductivity is an intrinsic property of materials. In this chapter we
may be ignored. shall consider the thermal conductivity of various materials, such as gases, liquids,
.l When a moving fluid at one temperature is in contact with a solid at a different
temperature, heat exchanges between the solid and the fluid by conduction at a
and solids, with the emphasis on solids, including not only dense solids but also
porous bulk materials.
'1
,1.
rate given by Fourier's law where k is the conductivity of the fluid, and JT/Jy
is the temperature gradient in the fluid normal to the wall at the fluid-solid
interface. If the details of the convection are known in a given situation, then we
I
1
To T1
on heat transfer has been devoted to the determination of h, since it enters into all t=0
su~nly raised
Small t Large t Steady-state
temperature
problems of heat transfer with convection. to em pera ture T 1
profile
I
Convection is usually classified as either a forced convection or free (natural)
Fig. 6.1 Build-up to steady-state temperature profile for a solid slab.
convection. When a pump or other mechanical device causes the fluid to move,
we call this process forced convection; when a fluid moves as a result of density
difference, then we speak of a free or natural convection. Thus, when a radiator heat is gradually transferred across the solid toward the colder surface, which is
heats the air which rises, displacing colder air in the upper part of a room, the fluid kept at T0 . A steady-state temperature distribution is attained when a constant
motion is by natural convection. rate of heat flow through the slab is required to maintain the temperature difference
In this section of the text, we shall examine the thermal conductivity of materials T1 - T0 . It is found that for sufficiently small temperature differences, the rate of
and the various modes of heat transfer applied to situations which metallurgists heat flow per unit area q is proportional to the temperature difference and inversely
are likely to encounter. proportional to the distance between the surfaces; hence,
These three relations form the components of the single equation Here the average distance moved in the y-direction between collisions is ji and
q = - kVT. (6.4) ji= 12, where Jc is the mean free path as defined by Eq. (1.7).
If the heat capacity per molecule is c, then the net flux of thermal energy
Note.the sii:iilarity between Eq. (6.2) for one-dimensional heat flux and Eq. (1.2) across the plane at y is given by the net difference between the energy of the mole-
for o~e-d1mens10nal momentum flux. In each case, the flux is proportional to the cules crossing they-plane in the positive and in the negative directions:
gradient of a potential variable, temperature, and velocity, respectively. The
qy = ZcTly-.ii - ZcTiy+.ii
mathematical similarity ends, however, when we compare the three components I
for heat flux versus the nine components that express momentum transfer in three = Zc~Tly-Ji - Tly+_v) (6.7)
dimensions. This is because energy is a scalar quantity whereas momentum is a
Equation (1.6) gives the expression for Z, the flux of molecules (number/sec-cm 2 )
vector quantity.
crossing the y-plane in either the positive or the negative direction; Eqs. (6.5) and
The units involved in Fourier's law of heat conduction are
(6.6) give the expressions for the temperatures, so that ultimately
I Engineering units Scientific units*
q = _ ncVJc(iJT) (6.8)
Btu/hr-ft 2 cal/sec-cm 2
r qx, qy, qz
T F or R 0
C or K
y 3 iJy
Denoting nc as C", the heat capacity per unit volume, and by comparing Eq. (6.8) to
x,y,z ft cm Eq. (6.2), we arrive at
k Btu/hr-ft F cal/sec-cm c
0
k = CJ'Jc. (6.9)
The thermal conductivity of a material reflects the relative ease or difficulty of 3
the transfer of energy through the material. This, in turn, depends on the bonding
This equation is basic for an understanding of the thermal conductivity of gases.
and structure of the material. When considering the thermal conductivity of most
In some cases, it can be extended to aid in theorizing the thermal conductivity of
materials used in engineering, bear in mind that the effective thermal conductivity
solids and liquids. For dilute gases with atoms assumed as rigid spheres, Eqs. (1.4)
depends on the interaction between the intrinsic thermal conductivities ofthephases
and (1.5) may be substituted for V and Jc in Eq. (6.9), with the result that
present and the mode of energy transfer between them. In the following sections,
we shall first consider the thermal conductivity of various pure phases, and then
k = __!_2 /Kfi ... - (6.10)
the effective thermal conductivity of some bulk materials. d V~
I I This result suggests that the thermal conductivity of gases does not depend on
* 1 caljsec-cm C = 241.9 Btu/hr-ft 0
F. pressure, but that it does depend on the square root of the temperature. For
-------rn.::nnart....-uiiuucnv1y-urgases-il'.'J.
Temperature, C
pressures to about 10 atm, this lack of dependence on pressure is essentially correct.
The thermal conductivity, however, varies with temperature more than predicted
by Eq. (6.10). H, x
In the case of polyatomic gases, Eucken 1 developed an equation for the thermal
conductivity of these gases at normal pressures:
_
k -11 (cp+--;:t
l.25R) (6.11)
""'
.::
tj
0.30
-g"
where M is the molecular weight and CP the heat capacity at constant pressure. 8
In the case of gas mixtures, we can estimate the thermal conductivity within a
few percent by
'"
"'
.<::
0.25
10
f-<
kmix = (6.12)
0.20 8
where X; is the mole fraction of component i having molecular weight Mi and
intrinsic thermal conductivity k;. A comparison 2 of observed conductivities with
calculated values, using Eq. (6.12), for binary gas mixtures involving air, CO, C0 2 ,
H 2 0, N 2 , N 2 0, NH 3, CH 4 , C 2 Hz, He, and Ar, indicates average discrepancies of 0.15 6
only 2.7 %, over the temperature range 273-353K. Tests at elevated temperatures
are not available, but the errors would probably be no larger. There are some far
more complex equations, which reduce the errors to about 1 %, but they are not
as easy to use. 4
Figure 6.2 gives the thermal conductivities of several common gases as a
function of temperature. The data are valid to at least 10 atm. For corrections at
higher pressures, the reader should consult references 34 below, and for more
detailed data, Tsederberg. 5
.I
i
1
A Eucken, Physik Z. 14, 324 (1913). Using Eq. (6.12) we get
! 2 L. Friend and S. Adler, article in Transport Properties in Gases, Northwestern University
Press, Evanston, Ill., 1958. k {2) 113 (0.295)(0.6) + (16) 113(0.127)(0.4) Btu
3 R. Bird, W. Stewart, and E. Lightfoot, Transport Phenomena, Wiley, New York.
mix = (0.6)(2)1/3 + (0.4)(16)1/3 ' - = 0.199 hr-ft op
4 J. 0. Hirschfelder, C. F. Curtiss, and R. B. Bird, Molecular Theory of Gases and Liquids,
the lattice is made up of independent oscillators, and that these oscillators are
simply considered to be elastic waves traveling in different directions with different
polarizations and wavelengths, not necessarily associated with the atoms them-
selves. This has led to his theory of specific heat (Cv oc T 3 at low temperatures)
which is more in agreement with experiments than Einstein's theory, and also
gives a more satisfactory picture of thermal conductivity of insulating solids. 0.5
0.002
Each lattice vibration (there is always a spectrum of vibrations) may be described
as a traveling wave carrying energy and obeying the laws of quantum mechanics.
By analogy with light theory, the waves in a crystal exhibit the attributes of
0.001
particles and are called phonons. 0.2
If the forces between atoms were perfectly harmonic, two phonons moving
through the crystal could collide and combine, with the resulting phonon having
the same total energy and momentum; energy transport would continue in the 0.0005
0.1"
direction of the resultant of the original phonons. If this were the only type of
phonon-phonon interaction (Normal or N-type), then the only limit to the mean
free-path of the phonons would be the dimensions of the piece of the material
0.05 0.0002
itself, since the resultant phonon acts in the same manner as the two original
jl I phonons as far as energy transport is concerned. Thermal conduction would thus
be extremely easy. But the forces between atoms are not in perfect harmony, and
there exists another type of collision between phonons, known as an Unklapp 0.0001
0 02 ~o_...c_::_...J....__.J__L__-;-1~
oo:;;:o:--..1__L___j_ __L----:2=-=o'-=o-=--o-1____J_ __J__ __J___J3000
(U-process )collision. In this process, energy is conserved, while phonon mGmentum
is not, and the result of such a collision is that a phonon moves in a direction opposite Temperature, F
to the resultant of the original phonons. For these processes, the mean free-path Fig. 6.3 Thermal conductivity of oxides and various insulating materials. (From A. Schack,
of phonons is extremely short, being of the order of the distance between atoms. Industrial Heat Transfer, Wiley, New York, 1965.)
Since the number of phonons increases with temperature, the number of the U-
processes also increases with temperature and the wavelength of the phonons A.Ph ... -
is proportional to 1/T. At room temperature and above, Cv for most materials is temperature. This is what happens in most electrically insulating substances, such
roughly constant, and if we use Eq. (6.9) to describe the thermal conductivity of a as the oxides shown in Fig. 6.3 (but not in the form of porous, bulk materials).
solid which conducts energy only by phonons, then k must decrease with increasing When heat enters a crystal, the crystal expands, and this thermal expansion of
L
the lattice is related to the bulk modulus B by a constant y, known as Gruneisen's Phonons are also scattered by differences in isotopic masses, chemical im-
constant purities, dislocations, and second phases. With these imperfections quantitative
')' = 2J!~L = 3~~a, (6.13)
prediction of thermal conductivity is exceedingly difficult. The only thing that we
can be sure about is that the less perfect the crystal is, the lower will be the thermal
where V = molar volume, a = linear thermal expansion coefficient, and Cv = conductivity due to phonons.
molar heat capacity. For most solids at ordinary temperatures, y ~ 2. Combining As we proceed from electrical insulators to conductors, we deal with materials
this result with Lindeman's theory of the melting of solids, Debye showed that with increasing concentrations of conduction electrons. The conduction electrons
in metals form an electron gas which obeys the laws of quantum mechanics. The
20Tmd 1
_
electronic contribution to the total heat capacity (per cm 3 ) of a metal is given by
/\ph - -2-' (6.14)
y T the expression
where Tm = melting point, T = absolute temperature, and d = crystal-lattice
n n.K~T
2
dimension, resulting in the previously mentioned intuitive relation between phonon cv,el = --'---'-- (6.15)
mean free path and temperature. Note that a high-melting-point material has a 2i;F
large value of A. at low temperatures, and therefore a large value of k at room Here i;F is the Fermi energy of the particular metal and n. is the number of free
temperature. We can observe this in the case of diamond, which has a thermal electrons per cm 3 . The Fermi energy is related to the average velocity of the electrons
conductivity at room temperature comparable to that of copper, one of the best by
conductors known. However, the conductivity of diamond drops very rapidly to
a small fraction of that of copper as the temperature increases.
(6.16)
At very low temperatures (T < ~ l0K), A.Ph approaches the dimensions of
the crystal, since U-processes diminish rapidly as the temperature approaches
absolute zero; consequently, the thermal conductivity rises rapidly. In this case, where VF = electron velocity at the Fermi surface, and m. = electron mass. Using
the larger the crystal, the higher the thermal conductivity, as illustrated in Fig. 6.4; Eq. (6.15) for the heat capacity and Eq. (6.16) for the Fermi energy, Eq. (6.9) becomes
this is an exception to Fourier's law.
(6.17)
~ LO c,<11'
u <o
" ~"'"'
~
.-2_ ~v where k. 1 is now the electronic contribution to the thermal conductivity. This
-;;;
u : " predicts that the electronic contribution in metals increases with temperature,
'()<$'
~-1 ':Q"" 0.5 '?~ provided that A.. 1does not decrease just as strongly with temperature.
u
To prove the theory that the electrons in a metal carry a major portion of the
:;I
'O thermal energy, the thermal and electrical conductivities should be proportionally
"u
0
related. This proportionality is known as the Wiedmann- Franz law and the
-;;; 0.3
constant of proportionality as the Lorentz number L:
"
.t:
---Solid ~ 0.20 ~
---Liquid .::::
;- e
-2.
;;:; "
.,,,_- 45 .,,,_-
1.00
b
0.18
:E
'O""
8" 40
-;;;
0.16
" ~1 0.80
Cu
t=:"
35
0.14
150 0.60
25
---"D"-Nickel (95 Ni-5 Mn)
0.10
ve-1. ve-
50 --- -----Al
w
020
20
:._\_61
~,.:,o
c,,-i
0.08
-o\\e
Fe \'-'"'
Electrical resistance alloys
15
Temperature, C Conversely, other oxides, for example, Si0 2, Al 20 3 , and MgO, never show
50
100
,., 300 500 700 900 1100 significant electronic conduction, and yet their thermal conductivities also increase
ff' Composition, wt % ;;
0
<.> with temperature at very high temperatures (see Fig. 6.3). This phenomenon has
-9 ~ c Si Mn Cr Ni Condition 0.20 ~
E. I .Pure iron Annealed E:<.> been explained on the basis of the ability of certain materials to transmit radiant
P5"' 45 2 0.23 0.11 0.63 - Annealed ;:::, energy (a photon contribution to thermal conductivity). This occurs when a material
_,,,_-
0.32 0.18 0.55 0.71 3.4 Annealed "'<.>
,., 3 0.18 ._,;_ is no longer opaque, but translucent to incident radiation. Without going into the
4 3.50 2.20 0.30 - 1.3 As cast
:E details of the interaction of radiation and solids, we can compute the radiant energy
<.> 40 5 0.17 0.25 13.00 0.1 Annealed
"'<:
't:I 0.68 0.37 19.10 8 .1 Quenched 0.16
conductivity k, from
0 1.30 1.20 15.20 26.9 16 uv 2 T 3
<.)
l
7
D.R. deHelar, Nucleonics 2, 92 (1963).
-~---------
i-nermarconoucav11.)'ornqu1us--Prt
"-----------------------'"'""J""..,I--~"'&&<:JI'-
k = 2.8K8 V, (
N
Vo
)2;3 , (6.21)
estimates if experimental data are not available. Figure 6.9 gives a summary of
the typical ranges of thermal conductivities for various classes of materials.
J_
8
9 J. R. Wilson, Structure of Liquid Metals and Alloys, Institute of Metals Report, 1967, page 482.
Bird, Stewart, and Lightfoot, ibid., page 260.
------__,.---------..----- ~------ ---------------w>s-o--.:r-..-_...,.-...-wo_.--.,.--..
-.....-::-~------
[;"<
---Experimental results, T= 100 C
"~
Solid metals and alloys I ~
a
10-1 ""
Crystal line
Molten metals, salts,
and slags
non-meta Ilic
solid
""'
10-2
10-2
0.4
10-s
10-3 ' - - - - - - - - + - - - - - - - - - - - - - - - - - - - - - - '
Fig. 6.9 Summary of the ranges of thermal conductivity for various classes of materials.
0 0.04 0.12 0.16
6.5 THERMAL CONDUCTIVITY OF BULK MATERIALS Volume fraction Al2 0 3
So far, we have presented equations, and looked at data describing the thermal Fig. 6.10 Thermal conductivity of Al-Al 2 0 3 (SAP) alloys at 100C. Similar results are
conductivity of individual phases. We have seen the difficulties of arriving at exact obtained on data available at 500C. (Data from D. Nobili and M.A. DeBacci, ibid.)
predictions of thermal conductivity for most solids, particularly complex alloys.
In most engineering situations, we are often faced with even more complex materials, Data for the thermal conductivity of SAP Al-Al 2 0 3 alloys 10 are plotted in
for example, porous bulk materials. The thermal conductivity of such materials Fig. 6.10, along with kmix calculated by using Eq. (6.23) and the data from Figs. 6.3
is certainly not equal to the intrinsic thermal conductivity of the solid involved. and 6.5. The agreement is less than satisfactory.
The question is: What value should be used? However, if we assume that the material behaves as if it were a series of plates
of Al and Al 2 0 3 normal to the direction of heat flow, then they would be equivalent
6.5.1 Two-phase mixtures to a series of resistors in an electric circuit, and
In nuclear and aerospace technology, alloys known as cermets have been evolved.
These are usually metallic alloys with a dispersion of a ceramic phase within the 1 ~ v.i
-=-+- (6.24)
matrix. The thermal conductivity of such materials is a function of the volume kmix kc kd
fraction of each phase. For up to one tenth volume fraction, Y,i, of a dispersed Using this equation, we obtain a much better agreement with the experimental
phase of spherical particles with intrinsic thermal conductivity, kd, it has been results. We should emphasize, however, that we lack the physical basis for this
suggested that the thermal conductivity of the mixture, kmix, may be found from assumption, and that the purpose in including it at this point is only to stress the
the Maxwell-Eucken equation, difficulty of predicting k for two-phase materials.
6.5.2 Porous materials
(6.22) Most ceramic materials of construction and some powder metallurgy products
have some, low, internal porosity w, as a result of having been sintered from powders.
The pores are generally isolated from one another. At temperatures up to and
If the thermal conductivity of the continuous phase, k 0 is much larger than kd, then slightly above the room temperature, the porosity has a thermal conductivity
kmix ~ kc l ( l-Y.i)
+ Y,i/ 2 ' (6.23)
i
10
D. Nobili and M.A. DeBacci, J. Nuclear Materials 18, 187 (1966).
l
~ 0.16 particle-to-particle thermal radiation also increases. Schotte 11 has developed a
u technique for predicting the conductivity of packed beds, in which the gas phase
"'
~ is continuous; it is outlined below.
-2-
-;;;
u
0.14 Taking the gas first, we can use the methods outlined in Section 6.2 to obtain
= 200c
~ kg, except inlhe following instance. When the effective pore dimensions are of the
~
same order of magnitude as the mean.free path of the gas, the true thermal con-
~u
0 ductivity of the gas decreases. This occurs when the particle diameter DP is about
-0
"'u0 1000 times the mean free path of the gas. (The mean free path of air at room
-;;; temperature and pressure is ~ 1o- 5 cm. At 2800F and 1 atm, it is ~ 10- 4 cm.
"'
.c:
Thus we are talking about situations where DP < 10- 2-10- 1cm.) Deissler and
E-< Eian 12 have developed a correlation for the break-away pressure Pb in lbf/ft2,
above which no correction for this effect is required:
T
Pb= (1.77 x 10 -21 ) - -, (6.26)
DPd 2
where T = the absolute temperature, R, DP = the average particle diameter, ft,
0
and d = the mean diameter of the gas molecules, ft. If the actual pressure is greater
than Pb, no correction to the value of kg read from Fig. 6.2 is needed, but if the
actual pressure is less than Pb calculated by Eq. (6.26), the thermal conductivity of
the gas phase should then be calculated according to the equation
ko
k = . g , (6.27)
g 1 + 2.03 x 10 -22( Cp/Cv ) (1 - W)( Tkg )
1 + Cp/Cv g W PDP d Cpl]
OL_~~~~-O~.l~~~~~0~.-2~~~~~0~.3~~~~~~0.4 where k~ is the uncon;ected value obtained from Fig. 6.2; all units for the variables
Porosity w are the same as for Eq. (6.26) with CP in Btu/lb?F and 17 in lb/ft-hr. It can readily
Fig. 6.11 Thermal conductivity of phosphor-bronze powder compacts as a function of be seen that kg decreases, at any given temperature, as DP decreases, and ifthe pores
porosity w. (From P. Grootenhuis, R. W. Powell, and R. P. Tye, Proc. Phys. Soc. 65, 502 are small enough, kg is effectively zero. This effect is illustrated in Fig. 6.3 by the
(1952).) data for the thermal conductivity of porous brick.
Having obtained kv either directly or by using Eq. (6.27) if required, we use
k., the intrinsic thermal conductivity of the solid phase, to calculate the ratio ks/ kg.
which is essentially zero. The effective thermal conductivity of the bulk material is Then, using- Fig. 6.12 we may find kb, the effective thermal conductivity of the
ofsen estimated to be packed bed, from the ratio kb/kg. This correlation (Fig. 6.12) has been developed
(6.25) from data on many packed beds, and is accurate as given at temperatures less than
200C. 13 However, Fig. 6.12 should not be used outside the limits 0.2 < w < 0.6
At larger porosities, the pores are not simple isolated shapes and the porosity and 1 < ks/kg < 6000. If ks/kg is greater than 6000, then there may be appreciable
becomes continuous. Conductivity at room temperature is thus even lower than contact conduction between particles which has not been accounted for in the
that predicted by Eq. (6.25). Figure 6.11 illustrates the effect of larger amounts
of porosity on the thermal conductivity of partially sintered powder metal com- 11
W. Schotte, A.I.Ch.E. Journal 6, 63 (1960).
pacts. 12
In general, the conductivity of a loose packing, or packed bed, is a function R. G. Deissler and C. S. Eian, NACA,.RM E52C05 (1952).
13
For another substantial review article, see R. Krupiczka: Analysis of thermal conductivity
of the thermal conductivity of the gas in the pores, the solid, the void fraction
of granular materials, Intl. Chem. Engr. 1, 122 (Jan. 1967). He arrives at a graph that is virtually
in the bed, and the temperature, because, as the temperature increases, the the same as that of Deissler and Eian.
..
- - - - - - - - - - - - - - ------------ --- - - -.-nenna1conoucuvrryorou11rmaTena1s--,;03
----T~
_,,,.,_,,,,. 6000
. where h, is the radiation heat-transfer coefficient between a particle and its neighbor,
given by
.,,...- I T3
; -~ h, = 0.692s , Btu/hr-ft 2 F.
108
"O
=1J=
u -
0
" "O
0"
u 0
Heres is the emissivity of the solid, and Tis the absolute temperature, R. Secondly,
] :;: 1000 there is radiation to and from the particle, in series with conduction through the
~c3 particle. This is written
Q = _ D2 1!:._( kh,DP )(dT).
8
(6.29)
4 ks + h,D
2
P dx P
When Q1 and Q 2 are added, one obtains the total heat flow
100
n -1-) (dT)
Q = -k, ( D 2 - - ,
P 4 1- w dx
where k, is the radiation contribution to the effective thermal conductivity of the bed:
IO
k, = (l/k:) ~ ~/k?) + wk?, (6.30)
and
k? = 0.692sDP(~:)
To evaluate the thermal conductivity of a packed bed, we use the correlation
in Fig. 6.12 (if necessary, use Eq. (6.27) to obtain kg), then add k, from Eq. (6.30) to
IO JOO 300 obtain the effective thermal conductivity, as illustrated in Example 6.2. The reader
Bed conductivity) = (1!.2) will note that even at high temperatures, it is possible for porosity to contribute
( Gas conduchv1ty kg
only slightly to k,, provided the pore size is quite small. This is generally true of
Fig. 6.12 Effect of porosity on packed bed thermal conductivity. (From Deissler and Eian, isolated pores as well as of pores in loose materials.
ibid.)
Example 6.2 Bearing in mind the importance of the effective thermal conductivity
of molding sand in relation to the solidification rate of castings, let us see if we can
f correlation of Fig. 6.12. However, this situation does not often arise, except possibly predict the thermal conductivity of silica sand molds as a function of temperature,
1I '
.I :
:, I
in compacts of powdered metals . usingifremethu&described above.
'
If the temperature is above 200C we must add another term to kb, found in
the same manner as above, in order to obtain the true effective thermal conductivity Solution. Assume that the material is entirely quartz (Si0 2 ) with thermal con-
.\I ductivity given in Fig. 6.3, and that the atmosphere is dry air. If the bulk porosity
of the bed. At higher temperatures, there is a contribution to the effective thermal
conductivity of the bed by radiation heat-transfer from particle to particle. This is 40 %, the grain size DP is O.Ql 5 in. (AFS 43), and the emissivity of quartz is as
given below, then we can proceed as outlined above.
contribution, for example, is important for 1-mm particles above 400C and for
0.1-mm particles above 1500C. Schotte considered the radiation from a plane on 1. Calculate Pb at 40F (500R)
one side of a spherical particle to a plane on the other side. First, there is a direct
,,' radiation heat-transfer across the void space past the particle, with heat-transfer
contribution:
I
i;'
;
! 1
Since atmospheric pressure = 2120 lb1 /ft 2 , then Pb is less than P. 1m, and therefore T, op 8 si02 kor k, + kb keff (Btu/hr-ft F)
we need to make no correction to the value read from Fig. 6.2 at this temperature.
40 0.82 0.00089 0.0009 0.156 0.157
Subsequent calculations show that this condition holds true up to 2100F. At
440 0.75 0.0047 0.0048 0.200 0.205
2340F (2800R), kg is calculated using Eq. (6.27). 0.68 0.0130 0.0131 0.229
840 0.242
2. Using the value of kg in the table below, we calculate the ratio ks/kg at 40F: 1240 0.58 0.0248 0.0245 0.248 0.273
1540 0.49 0.0339 . 0.0337 0.273 0.307
ks/kg = 5.0/0.013 = 384. 2040 0.32 0.0400 0.0398 0.396 0.436
2340 0.27 0.0514 0.0512 0.452 0.503
Then, from Fig. 6.12, we find that for w = 0.4, kb/kg~ 12, or, since kg = 0.013,
kb = 0.156 Btu/hr-ft F. 0 The results of these calculations are plotted in Fig. 6.13. Some data from
This operation is carried out at each temperature: experimental studies are compared with the calculated values. The results are in
surprisingly good agreement, and the calculations are probably as good as the
T, op kg ks/kg kb/kg kb Btu/hr-ft F measured values, considering the experimental difficulties involved.
40 0.013 384 12.0 0.156 For another interesting application of this method in analyzing a metallurgical
440 0.021 167 9.5 0.200
840 0.029 90 7.9 0.229 Temperature, C
~''11
On the other hand, at 2340F:
0.2
2
= 1.0 - 0.4 ( 4)(0 0515 )
.
1: k, (1/15.0) + (1/0.0515) + O. . 0.0
400 800 1200 1600 2000 2400
'
]
': I ,I = 0.0512 Btu/hr-ft F. Temperature, F
'11 Fig. 6.13 Calculated and experimental thermal conductivity of silica molding sand. (Experi-
These values are added to the kb values, previously calculated, to obtain the final mental values from L. F. Lucks, C. L. Linebrink, and K. L. Johnson, Trans. A.F.S. 55, 62
. i Ir (1947).)
result.
! I, .,, ',!'
I'.,
!
---'1= . .
7
PROBLEMS
6.1 In the same system described in Problem 1.2, the temperature profile at x = x 1 is given by
HEAT TRANSFER AND THE
T = 6 sin (n/2)y, 0 ~ y ~ 1, ENERGY EQUATION
where T is measured in F and y in ft. Find the heat flux through the wall at that point. (The
thermal conductivity of water at 100F is 0.36 Btu/hr-ft 2 F, and the specific heat is 1 Btu/lbm-F).
6.2 Determine the thermal conductivity of a test panel 6 in. x 6 in. and tin. thick, if during
a two-hour period 80 Btu are conducted through the panel when the temperature of the two We have designed this chapter to introduce the reader to three interwoven topics.
faces are 67F and 79F. First, we develop differential equations in terms of temperature in space (and with
6.3 At steady state, the temperature profile in a laminated system appears thus :
time if transient conditions apply) for several simple physical problems, by writing
energy balances for suitable unit volumes. In order to obtain useful solutions to the
Material I Material II problems, we integrate the differential equations to ascertain the temperature and
500F
................ arbitrary constants, and then apply boundary and initial conditions to obtain the
........... particular solution. The general procedure is conceptually identical to that followed
...........
............ in Chapter 2 for obtaining the velocity profiles .
------- 100F
Second, several of the examples to be discussed concern heat transfer to and
from moving fluids; in this regard, we deal only with laminar convection, but this
enables the reader to become involved in the fundamentals of heat transfer with
- - - - - 1.5 ft - - - - - - - 1 . 0 ft convection.
Third, we bring to the reader's attention more general forms of the equation of
Determine the thermal conductivity of II if the steady-state heat flux is 4000 Btu/hr-ft and
2
energy, leading to Tables 7.2-7.4 which may be used in a manner similar to the
the conductivity of I is 30 Btu/hr-ft F. general momentum equations given in Chapter 2.
6.4
a) The thermal conductivity of helium at 200F and 1 atm is reported to be 0.098 Btu/hr-ft F. 7.1 HEAT TRANSFER WITH FORCED CONVECTION IN A TUBE
What is helium's thermal conductivity (in the same units) at 1000F?
Consider a fluid in laminar flow in a circular tube of radius R, as depicted in Fig. 7.1.
b) Estimate the thermal conductivity (in Btu/hr-ft F) of carbon dioxide at 2000F. If the tube and the fluid exchange heat, then clearly the fluid's temperature is a
6.5 Show that Fourier's law can be written (for constant pCP) as function of both the r- and z-directions. A suitable unit volume is a ring-shaped
element, Ar thick and Az high. Energy enters and leaves this ring by thermal
d
q = -a-(pC T) conduction; also, a unit mass of fluid, which enters with an enthalpy corresponding
y dy p
to its temperature (sensible heat), must leave the ring with a different enthalpy.
for one-dimensional heat flow. In addition, show that Newton's law, for constant p, is Let us now develop the energy balance for the unit volume.
ii.! '
~
!
:1
i iI <yx = - v dy
d
(pvxJ.
! I
Rate of energy in by conduction across surface at r 2nr Azq,I,
I
! I iI .,1
J
Discuss the analogies between the fluxes, constants, and gradients as they appear in these Rate of energy out by conduction across surface at r + Ar 2n(r + Ar) Azq,lr+M
' i!
j equations.
I Rate of energy in by conduction across surface at z 2nr Arqzlz
14 J. H. Downing, Electric Furnace Proceedings, A.I.M.E. 26, 81 (1968). Energy out by conduction across surface at z + Az 2nr Arqzlz+L\z
207
/.~-----
-------------------- ---e';)'----.,--.......... .
Substituting Eqs. (7.3) and (7.4) into Eq. (7.2) yields an energy equation written in
terms of temperature:
oT k [1
0 ( OT)
Vzfh= pCP-;: Or ra;: + 0Z2 .
2
o TJ (7.5)
We can further simplify the energy balance (Eq. 7.5), since, except for the very
slow flow of liquid metals, the term (k/pCP) o 2T/oz 2 is negligible even though
vz(oT/oz) is not. With this assumption, Eq. (7.5) reduces to
Equation (7.6) contains vz, the factor that ties together heat transfer and convection.
For the purpose of this discussion, consider fully developed laminar flow; the
velocity distribution is therefore parabolic, and, as previously derived, is given by
Eqs. (2.3 l) and (2.33):
Fig. 7.1 Elemental circular ring used to develop the differential energy balance for laminar
By including this velocity distribution, Eq. (7.6) becomes
tube flow.
Energy in due to fluid flow (sensible heat) across surface at z pvz2nr ArHl 2
2r:[1 - (!._) ]0T = ~[~ ~(r
R
2
Oz pCP r or
OT)]
or
(7.7)
Here we consider the special case where a fully developed temperature profile
Energy out due to fluid flow across surface at z + Az
exists. For any set of boundary conditions, a fully developed temperature profile
Here H is the enthalpy per unit mass, and vz is the velocity in the z-direction. At exists when (TR - T)/(TR - Tm) is a unique function ofr/R, independent of z. Then
steady state, the energy balance requires equal inputs and outputs. Ifwe divide all
(7.8)
terms by 2n Ar Az, we obtain
or
(7.1)
~(TR - T) = O (7.9)
OZ TR - Tm '
Now Ar and Az are allowed to approach zero, where TR = temperature of fluid at the wall, and Tm= mean temperature of fluid.
A fully developed temperature profile is analogous to fully developed flow. This
o(rq,) oqz oH _ O (7.2) is exemplified by Fig. 7.2; where the liquid flowing in the z-direction encounters
or +r oz + rpvz oz - . the heated section of the tube. Over a finite interval downstream from this point,
the temperature profile changes from uniform to fully developed.
If cp is the heat capacity, then For a fully developed temperature profile, an important corollary arises;
namely, the heat-transfer coefficient is uniform along the pipe. We realize this by
oH oT (7.3) employing the definition of the heat-transfer coefficient based on the mean tempera-
-=C-
oz p oz ture of the fluid:
Also k 0 (TR-T) - (7.10)
q, = - k(oT/or) and q2 = - k(oT/oz). (7.4 a, b) R o(r/R) TR - Tm r=R
--------------e;r----.----
~
B.C.1 at r = 0,
Integrating, we get
Entering
1_ ~(!_)
2
..
uniform
tern pera ture 2v(oTm)!,_[ ] = ~ oT. (7.15)
Power for'. . !
profile zoz 2 . 2 R pCP or
resistance . , , 9
heating . : A second integration with
WH'---0:,!11111!'
'\~ ~:;:Developing
B.C. 2 at r = R,
,~: :,yrofile
finally results in the temperature distribution
T - T= (V,pCP)(oTm)(3R4
R 8R 2 k oz
- 4r2R2 + r4) . (7.16)
Having obtained the temperature profile, we can evaluate h. From Eq. (7.10),
(TR - Tm) and q 0 must then be evaluated. First, we find (TR - Tm) by performing
the integration:
Fig. 7.2 Heating a fluid in a tube showing the development of the temperature profile.
r Vz(TR - T)2nr dr
TR - Tm = -"--"-o-r_R_ _ _ __ (7.17)
Because the derivative in Eq. (7.10) has a unique value at the wall, independent Jo vz2nr dr
of z, h is therefore uniform along the pipe under the fully developed temperature
conditions. Second, we determine q 0 by evaluating the gradient at the wall using Eq. (7.16):
Now consider the case where q 0 is uniform. This represents a uniform heat
flux at the wall, and could be physically obtained by using an electric heater,
qo = -k(oT) . (7.18)
depicted in Fig. 7.2. Further, since hand q0 are constant, Eq. (7.10) specifies that or r=R
TR - Tm is constant, and
When these operations have been carried out, we can determine the heat-transfer
(7.11) coefficient. The final result is
(Note that TR and Tm themselves are not constants.) Now expand Eq. (7.9) in a
general sense where each quantity varies as follows,
h = 2.18 ~ (7.19)
This Nusselt number developed here is for fully developed flow and uniform Adding all these quantities, dividing through by Ax Ay, and taking the limits
heat flux with parabolic velocity profile. It is subscripted with oo because it as Ax ~ 0 and Ay ~ 0, we obtain
represents a limiting case. Many other situations have been analyzed, some of
oqx + oqy + o(pvxH) + o(pvyH) = O.
which are -given in Table 7.1. (7.22)
ox oy ox oy
For constant density and conductivity, Eq. (7.22) becomes
7.2 HEAT TRANSFER WITH LAMINAR FORCED CONVECTION OVER A FLAT
PLATE - 2 2
oDx oDy) ( oH oH) (o T o T) (7.23)
pH (-;;; + oy + p Dx OX + Dy oy = k OX2 + oy2 .
In Chapter 2, the velocity distribution, within the boundary layer, of a fluid flowing
past a flat plate was determined by two methods-the exact solution, and the Since continuity requires that (oDx/ox) + (oDy/oy) = 0 and dH = CP dT, we finally
approximate integral solution. Here we consider the case of a plate at a higher obtain
temperature than the fluid, the plate serving to heat the fluid. Just as a velocity 2 2
profile continually changes with distance from the leading edge,,and results in a oT oT) (o T o T) (7.24)
pCP ( Vx ox + Dy oy = k OX2 + oy2 .
momentum boundary layer which increases in tnickness, there is also a changing
temperature profile and development of a thermal boundary layer when heat The temperature gradient in the y-direction is much steeper than that in the
transfer is involved. We depict this situation along with a unit element in Fig. 7.3. x-direction, therefore the x-directed second derivative term (conduction term) may
be neglected. Then Eq. (7.24) simplifies to
oT oT o 2T
Dx OX + Dy oy = IY. oy2 . (7.25)
Here
k
IY.=--
pCP
Fig. 7.3 Development of the thermal boundary layer and the temperature distribution over a which is called the thermal diffusivity. The thermal diffusivity has the units oflength
flat plate. squared per time (L 2 /t), which are also the units of the kinematic viscosity. In
--~-
. ---~
r-------------
! .'
addition, the momentum boundary-layer equation, developed in Chapter 2, is t Knowing the temperature profile, we can determine the heat-transfer coeffi-
cient. From the results given in Fig. 7.4, the local heat transfer is
(2.80)
(7.26)
Equations (7.25) and (2.80) are analogous. If v = a, and if the velocity and thermal
boundary conditions are similar, then the temperature and velocity profiles are
exactly identical, and the thermal boundary layer br equals the momentum Pr 2 0.6,
boundary layer /J.
or, in terms of dimensionless groups,
i 7.2.1 Exact solution Nux = 0.332 Pr 0 343 Re} 5 , (7.27)
The method of solution for Eq. (7.25) parallels that for the velocity distribution, where
as given in Section 2.7.1. For the case ofa uniform-temperature plate, the following
boundary conditions apply Nux =hxx,
k
B.C. 1 at y = 0, which is called the local Nusselt number. If we wish to know the average heat-
transfer coefficient, then we can find it by averaging hx from x = 0 to x = L:
B.C. 2 at y = oo,
L
h = ~J hx dx = 0.664k Pr
0
0 343
J[,,
or
/JT 5.0
Fig. 7.4 Dimensionless temperature profiles in the laminar boundary layer over a flat plate (7.29)
for various Pr. (From E. Z. Pohlhausen, Z.f. angew. Math. u. Mech. 1, 115 (1921).) ~=~,
0.564
,,
"
Nux = jRex Pr (
1 + 0.90
JPrPr ) (7.30)
For the other boundary condition of uniform heat flux at the wall, we present
x x +fix
these results :2
Fig. 7.5 Integrated enthalpy quantities in the boundary layer over a flat plate.
Pr> 0.5, Nux = 0.458 Pr 0 343 Fe:"; (7.31)
We write the respective total enthalpies I that accompany Wx, WxHx and W,, and
0.880
)
0.006 ::;; Pr ::;; 0.03, Nux = jRex Pr ( JPr
1 + 1.317 Pr
(7.32) enter them into the energy balance. We neglect conduction in the x-direction.
assume that br < b. We can use the same approach with a slight modification for 0/ 0
br > b (liquid metals). Following a similar procedure used for the analogous
momentum balance in Chapter 2, the mass of material entering across the surface I,
at xis l / l
Wx = f
l
pvxdy, (2.98)
= JHfv:~Y + d~(f Hpvxdy) Ax, (7.34)
0 / 0
/
0
and the mass leaving is
l where H == enthalpy per unit mass.
wx+L\x = wx + :Xu 0
PVx dy) Ax. The integrals in Eq. (7.34) are then split, that is,
br
Continuity requires that
l f =f~J
d~(f pvx dy) Ax.
0 0
W, = (2.100)
and then simplification yields
0
1
E. M. Sparrow and J. L. Gregg, J. Aero. Sc. 24, 852 (1957). (7.35)
2
R. J. Nickerson and H.P. Smith, as reported in Rohsenow and Choi, ibid., page 149.
- - -H-e-at-transter-w1m-1am1narmrce-dconve-ctron-over-a-nat-p1ate--:i:1Y
T
r-;.----~
Since H - H 00 = Cp(T - T 00 ), we obtain, for constant p and CP, , - Frnm Eq. (7.371 the loeal heat-trnmfec eodlicient beeome
or 1~,
h ~ -k(oT/oy)y=o = ~ !!___ (7.4 3)
:x(f 0
(T00 - T)vxdy) = a(~:L=o (7.36)
1
x (T0 - T J 2 bT0
b; = l.02~Frf-(:0 t
4 135F, the kinematic viscosity is 0.725 ft2 /hr. Then,
V00 x 50 ft 0.5 ft hr 3600 sec = OOO
124
For the specific case of no unheated length, that is, the entire plate is maintained v sec 0.725 ft2 1 hr ' '
at T0 , then x 0 = 0, and the ratio of boundary layers is and
1
-- -=------------...........,.--- --------- ----------
. .lf!P.
or
b = 0.0852 in.
For our purposes, Eq. (7.42) is sufficient to give br. Therefore
br _ _1_ p -1/3
b - 1.026 r
The Prandtl number for air, evaluated at 135F, is 0.703. Then,
7.10 x 10- 3 -3
{JT = 3~
(1.026)~ 0.703
= 7.75 X lQ ft,
or
f>T = 0.0930 in.
b) Eq. (7.28) is used for the average heat-transfer coefficient which applies to the
whole plate. This equation is often applied with a slight modification in the
exponent of Pr; the value 0.343 is replaced by l
NuL = 0.664 Pr 113 Rell 2 , Fig. 7.6 Thermal and momentum boundary layers for vertical plate natural convection.
NuL = 0.664~J248,000 = 295. therefore we write the momentum equation for the x-component only. In the
The thermal conductivity for air at 135F is 0.017 Btu/hr-ft F. Then deri~ation of the momentum equation for steady forced convection over a flat plate,
we ignore the gravity force, and no pressure gradient is involved. We cannot ignore
k NuL = (0.017)
h= L -1- (295) = 5.02 Btu/hr-ft 2 F, these forces in free convection, and therefore the momentum balance for the present
case contains these terms
and finally ovx ovx 1 oP o 2vx
Q = h(T00 - T0 )A = (5.02)(200 - 70)(1) = 653 Btu/hr.
Vx OX + Vy oy -p OX + V oy2 - gx. (7.46)
In they-direction, oP/oy = 0, or at any x, the pressure is uniform regardless of
distance from the wall. The pressure gradient in the x-direction is due to the weight
7.3 HEAT TRANSFER WITH NATURAL CONVECTION
of the fluid itself. Therefore, it can be expressed as oP/ox = - p 00 gx, so that in
In Sections 7.1 and 7.2 we considered two examples of heat transfer with forced Eq. (7.46) the pressure gradient term becomes
convection. In forced convection, the fluid flows as a result of some external forces, 1 oP 1
resulting in a known velocity distribution which can be entered into the energy - - ~ - gx = - gx(Poo - p) = /3gx(T - T 00
). (7.47)
p ux p
equation. The situation is more complex in problems of heat transfer with natural
convection. The energy equation still contains terms with velocities, but these The volume expansion coefficient f3 is defined as
velocities are not known a priori to solving the energy equation. The momentum
balance depends upon the temperature profile which, of course, cannot be solved
until the temperature distribution has been established. Hence, the velocity and
/3 = -~(op)
p oT P
.
temperature distributions cannot be treated as separate problems; the temperature The momentum balance in the x-direction then becomes
distribution, in effect, produces the velocity distribution by causing density 2
differences within the fluid. ovx ovx 0 Vx /3 8)
vx OX + Vy oy = v oy2 + gx (T - Too), (7.4
Consider specifically the vertical surface in Fig. 7.6; the surface is at T0 , and it
heats the neighboring fluid whose bulk temperature is T00 In this situation, the which is identical with the equation of the forced convection boundary-layer,
velocity component Vy is quite small; the fluid moves almost entirely upward, and except for the buoyancy term. Equation (7.48) immediately shows that the
I
- - - - - - - - - - - --------
momentum balance must be coupled to an appropriate energy equation, in order Equation (7.52) if rewritten without all the K's would of course be valid, since it
to treat the buoyancy term. would transform back to Eq. (7.48). Hence,
The energy equation applied to the control volume Ax Ay in this case is
identical to that for flow over a flat plate:
(7.53)
aT aT a 2T
Vx-0 +Vy-(} = a-0 2. (7.49) and therefore
x y y
VzVz
This equation is coupled to Eq. (7.48) by the presence of the velocity terms. The
Lz2
mathematical task at hand is, therefore, to solve coupled Eqs. (7.48) and (7.49), for
the boundary conditions: (7.54 a, b, c)
and
B.C. 1 at y = 0, Vx =Vy= 0,
B.C. 2 at y = oo, Vx =Vy= 0,
If we combine Eqs. (7.54 a) and (7.54 b) we get
Analytical solutions of such coupled differential equations being beyond the scope V1L1 VzL2
of this text, we simply present the results. First, however, let us examine a dimen- --=--, (7.55)
V1 Vz
sional analysis approach in order to bring forth pertinent dimensionless groups.
The problem is to determine the conditions for which the velocity profile in a which are Reynolds numbers. The combination ofEqs. (7.54 b) and (7.54 c), yields
natural convection situation is similar to the velocity profile in another natural
gi/31(To - Too)1Li gzf32(To - T00 )2L~
convection situation. Both systems have the same boundary conditions, i.e., (7.56)
velocity is zero at the surface and within the bulk fluid removed from the surface.
Now employ the dynamic similarity argument introduced in Section 3.1. The group of variables represented in Eq. (7.56)could be considered as an important
First, Eq. (7.48) is written for system 1: dimensionless group, but by reflecting on the physical aspects of the problems, we
realize that the velocity of the fluid is not an independent quantity, but that it
(7.50) rather depends on the buoyant force. Hence the v's are eliminated from Eq. (7.56),
and substituting their equivalents from Eq. (7.55), we obtain
System 2 is related to system 1 by geometrical and dynamic similarities expressed gi/31(To - Too)1LI gzf32(To - T00 )zL~
(7.57)
by the ratios:
Lz This dimensionless group of variables which is important in natural convection
K L -= X1 -- Y1 K =~ problems is called the Grashof number, Gr. When the buoyancy is the only driving
Xz Y2 Lz v Vz
I
force for convection, the velocity profile is determined entirely by the quantities in
K = Vx1 = Vy1 V1 K = gi the Grashof number, and the Reynolds number is superfluous.
'V
(I'
v-
Vxz Vy2 Vz g gz
(7.51)
Regarding heat transfer with natural convection, recall that for forced con-
i'
vection, as discussed in Section. 7.2, the Nusselt number was correlated in the
K p-- /31 (T - T00 ) 1 (To - T00 )1
general form
/32 Kr= (T - T00 )z (To - T00 )z
Nu = f(Pr, Re), forced convection.
Now, we replace vx 1 , x 1 , v 1 , vy 1 , g 1 , etc. in Eq. (7.50) by their equivalents in Eq.
(7.51); then we write Eq. (7.48) for system 2: Correspondingly then, for natural convection, the Nusselt number is correlated as
Nu = f(Pr, Gr), natural convection.
(7.52) Returning to the complete solution of Eqs. (7.48) and (7.49), we present the
results for velocity and temperature distributions (see Fig. 7.7). The curves show
. --.-.~-~---
.,. . ,,~~-------
><(I
~~1~8
llo'0.6
N
l::).
'
I
Example 7.1.
'
long x 1 ft wide hung vertically in air at 70F. Contrast the results with those of
0.5
Solution. Equation (7.58) should be integrated to obtain the average heat-transfer
coefficient which can be applied to the whole plate.
In Eq. (7.58), since hx varies as x- 114 , then the average h .equals 1'hx. Hence
NuL defined as hL/k is
0.902 Pr 112
(7.59)
(0.861 + Pr) 114
or
NuL = 0.902
1 GrL Pr
4(0.861 +
2
Pr). (7.59a)
Next, we calculate the product GrL Pr to test for laminar flow conditions
4 5 6 GrL Pr = (1.82 x 10 8 )(0.703) = 1.28 x 10 8 .
( b) (Grx)\4
4 . x
,!'_
Since it is between 10 4 and 10 10 , laminar flow holds, and Eq. (7.59) is valid. When
Fig. 7.7 Laminar natural convection for a vertical plate. (a) Dimensionless velocity profiles. we substitute values of GrL and Pr into Eq. (7.59a),
(b) Dimensionless temperature profiles. (Calculated by S. Ostrach, Nat. Advisory Comm.
NuL = 82.9,
Aeronaut. Tech. Note 2635, Feb. 1952, as presented in W. M. Rohsenow and H. Y. Choi, Heat,
Mass, and Momentum Transfer, Prentice-Hall, Englewood Cliffs, New Jersey, 1961, pages from which
155-159.)
k = (82.9) (0.017)
h = (82.9) L - - = 1.41 Btu/hr-ft 2 F.
1
that for Pr ~ 1, by ~ b, but for Pr > 1, by < b. For liquid metals, therefore, by
Finally, we evaluate the rate of heat transfer Q.
is about equal to b in free convection as contrasted to forced convection in which
by b. Corresponding to the temperature profile, as shown in Fig. 7.7b, the results Q = h(T00 - T0 )A = (1.41)(200 - 70)(1) = 183 Btu/hr.
for the local Nusselt number correlation are
For Example 7.1, Q was 653 Btu/hr, that is, the rate of heat transfer for forced
0.676 Pr 112 convection is considerably higher. This is the usual case.
(7.58)
(0.861 + Pr) 114
It is instructive to look at special forms of Eq. (7.59a). First, if Pr = 0.702, then
Equation (7.58) applies for a wide range of Pr numbers (0.00835 ~ Pr ~ 1000), and it reduces to
has been shown to apply for laminar flow conditions, 10 4 < Grx Pr< 10 10 .
.(7.59b)
Example 7.2 Calculate the initial heat-transfer rate from a plate at 200',F, 1 ft
It so happens that for many gases, including air, 0 2 , N 2 , CO, He (and other inert Rate of energy in by conduction across surface at r 2nrlq,/,
gases), H 2 and C0 2 , Pr is very close to 0. 7, and practically constant for temperatures Rate of energy out by conduction across surface at r 2nrlq,lr+M
even as high as 3000F. Thus, we can apply Eq. (7.59b) directly to gases.
Second, if Pr--+ 0, then Eq. (7.59a) reduces to At steady state, these are the only terms that contribute to the energy balance.
Thus
4~
NuL = o.936v~ 2nlrq,/r+dr. - 2nlrq,/, = 0.
Therefore, a somewhat more approximate, but still adequate, form for most If we divide all terms by 2nl !J.r, and take the limit as !J.r approaches zero, we
calculations involving natural convection of liquid metals is obtain
NuL =
~ JGrL 4Pr 2
(7.59c)
Jim (rq,)/r+dr - (rq,)/, = 0
M-0 /J.r '
or
d(rq,)
-;i;:- = 0. (7.60)
-k dT = C 1 . (7.62)
dr r
Integrating once again, we find for constant thermal conductivity
C1 C2
T= --Inr - - (7.63)
k k
By absorbing k in new constants, Eq. (7.63) simplifies even more
Fig. 7.8 Heat conduction through a solid cylindrical wall. The shaded area depicts the unit
T= C 3 lnr + C4 . (7.64)
volume.
The boundary conditions under consideration are
In this section we consider the problem of heat conduction through the wall of a B.C.2 at r = r2 , T= T2
hollow solid cylinder. Figure 7.8 depicts the situation, and also locates a suit-
Determination of the constants using the boundary conditions yields the
able unit volume with a thickness !J.r. From a practical point of view, we may
temperature distribution
visualize the system as a long cylindrical shaped furnace, and it is desirable to
calculate the heat loss through the walls to the surroundings. Suppose the cylinder In (r/r 2 )
(7.65)
is long enough so that end effects are negligible; in addition, the system is at steady In (r ifr 2 )
state, so that both the inside and outside surfaces of the wall are at some fixed
temperatures, T1 and T2 , respectively. For such a system, we develop the energy Determining the heat flux (Btu/hr-ft 2 ), we note that it is not constant, as we
balance. have mentioned.
J I
~
rnegenerarenergy-equanon-ZZ9
- - - - - - - - - - c.., - - . - - -
2 (7.70)
In (r 2 /ri) = ( 0.2)
- (0.550) = 0.275, +Ax Ay{vz(pU + !pv 2 )lz - Vz(pU + !pv )lz+&z}.
I I 0.4
! In a similar manner, the net rate of energy in by conduction is
so that
Ay Az{qxlx - qxlx+&x} +Ax Az{qyly - qyly+&y} +Ax Ay{qzlz - qzlz+&J (7.71)
r 2 = 2.64 in.
----:.-11~~-a1-'Cl'.1~i;..,--~ua1u11--0k:n,------
The work done by the fluid consists of work against gravity, work against If all the terms involving (p U + pv2) are expanded and combined, we can obtain
pressure, and work against viscous forces. The rate of doing work against the three
components of gravity is
a
p [ot (U
a (U + v2) + Vy oyo (U + v2) +. Dz oy
+ v2) + vx OX o (U + v2) J
(7.72)
(7.78)
The rate of doing work against the pressure at the six faces of the unit volume
lS
Continuity (Eq. A, Table 2.1) requires that the second term in the above expression
AyAz{(Pvx)lx+L\x - (Pvx)lx} + Ax Az{(Pvy)ly+L\y - (Pvy)ly} is zero; the remaining term in the expression is the substantial derivative of
+ AxAy{(Pvz)lz+L\z - (Pvz)lz} (7.73) (U + v2 ), so that we may write Eq. (7.77) as
Similar expressions may be written for the work against the y- and z-directed
viscous forces
Ay Az{rxyvylx+&x - rxyvylx} +Ax Az{ryyvyly+&y - ryyv)y}
(7.75)
(7.79)
and
Ay Az{ rxzvzlx+&x - rxzvzlx} + Ax Az{ ryzvzly+&y - ryzvzly} For most engineering problems involving heat flow, it is more convenient to
have the equation ofenergy in terms of heat capacity than in terms of internal energy.
+Ax Ay{rzzvzlz+L\z - !zzvzlz} (7.76)
The manipulation that leads to the development of the following equation is rather
lengthy, and is not given here, but with the aid of the conservation of momentum
Substituting all these expressions into Eq. (7.68), dividing by Ax Ay Az, and
we can write Eq. (7.79): '
taking the limit as Ax, Ay, and Az approach zero, we obtain one form of the energy
equation
(7.80)
!!_ (pU + pv2) =
at
-(!.-ax
Dx (pU + pv2) + aa
y
vy(pU + pv2) + aa
z
Vz(pU
2
+ pv ))
where the quantity <I> is known as the dissipation function,
aqx
- ( ~ + aqy
ay +
oqz)
az- + p(vxgx + vygy + vzgz )
<I>=
2 [(~;r + (~;r + (~~rJ
~(ovx + OVy +
2
- ovz) (7.81)
3 ax ay oz
l
.~--~- --- ------ -- _,, ____ ----1neem:rgrequaTion-incurvmnearcimr01nates Z33
-------o-~--
DU DH 1 DP P Dp vz =
az az az
-+--+- (7.88)
- = - - - - + -2 - , (7.82) - ax 2 ay 2 az 2
Dt Dt p Dt p Dt
lI
')'
:;
equation in terms of temperature:
, II v Dt ax ax ay ay az t qy = -k-
ay (B) q9 = -k-- (E) q9 = -k--
r ae
(H)
. 1, r ae
If
11 Simplifications ofEq. (7.87) are usually referred to. For example, if k is independent ar ar 1 ar
t '.r i:
qz = -k az (C) qz = -k;;; (F) q =
4>
-k ---
r sine 8</>
(I)
I of the space coordinates, we write
DT DP * This table and the following two tables are from R. B. Bird, W. E. Stewart, and
pCv - = kV 2 T + T/3 -D + 17<1>, E. N. Lightfoot, Transport Phenomena, Wiley, New York, 1960, pages 317-319.
Dt t
-------------------------------------15-;T~-..,.........-..~--...... .,...........-.,,""'..... -Q'"'""-.;;::r;;J' _ _ _ _ __
Table 7.4 The equation of energy in terms of energy and momentum fluxes
Table 7.5 The equation of energy in terms of the transport properties (for Newtonian fluids
of constant p, and note that the constancy pimplies that
17, k; of Cv = Cp)
Rectangular coordinates
ar ar ar ar)
pCv (at+ Vx ax +Vy ay + Vza; = -
[aq aq aq
a:+ a:+ a:
J Rectangular coordinates
Cylindrical coordinates
Cylindrical coordinates
ao i
8
+ 211
+ 'l:zz avz
az }- {'l:,9 [rara(v-; ) + -;:1 av,]
ao + 'l:rz (avz av,)
8
a,: + a;
avz+av,)i
+ (-
ar- +
az
[1--+r--
av, a(v )Ji} .
r ao ar r
8
(B)
1
+ 'l:9z ( -;:
avz av .
8 )}
ao +a; (B)
Spherical coordinates
Spherical coordinates
ar v -
ar v ar - v<t> -ar) [ 1 a ( i ar)
(-+
8
ar ar v ar v<t> ar)
8
pCv (at+ v,a,: + 7 ao + rsin 0 a<f; = -
[1 a
-;:z ar (riq,)
pC
v at r ar +r-ao- +rsin -
(} a<1>
= k -2 - r -
r ar ar
+ _1_ ~(sin(} ar) + _2 1 _ a r]
2
!i
2
.j sin 0 a ( v<t> ) 1 av 8] } (C)
Ii1i + [ -r- ao sin (} + rsin (} a<f; .
(C)
it
Note: The terms contained in braces { } are associated with viscous dissipation and may usually be
Nate: The terms contained in braces { } are associated with viscous dissipation and may usually be
neglected, except for systems with large velocity gradients.
j;' neglected, except for systems with large velocity gradients.
11,
----------<J-,.------~
-----------------b-~r "'I--~ ....
in which
7.2 Utilize the approximate integral technique and develop an expression for the local
11k
Nusselt number for forced convection past a flat plate. Make the following assumptions: /3 =. p 2 Cpgb.
a) bT 5, b.
b) The velocity profile is linear. d) Write the boundary conditions for such short contact times.
c) The temperature profile is linear.
e) Rewrite the differential equation (part c) and the boundary conditions in terms of the
J 7.3 A liquid film at T0 flows down a vertical wall at a higher temperature T,.. Estimate the following reduced variables
~. rate of heat transfer from the wall to the liquid for such contact times that the liquid temperature T- T0
- changes appreciably only in the immediate vicinity of the wall. 8=--
T,. - To
y=y/~.
Integrate once to obtain
(C is an integration constant). Integrate a second time and use the boundary conditions
to obtain
6 = film thickness (J =
1
r(!)
Je -y3 d
y,
(n > 0).
0
f) Show that the average heat flux to the fluid (for length of solid L) is
a) Show that the energy equation can be written (state assumptions). (9/JL)1;3
iily=o = ~k(T,. - To) r(!)
ar air
pC v - = k - 7.4 Calculate
p z oz ov2 a) the rate of heat flow Q (Btu/hr) across a 2 in. thick slab of stainless steel for the following
situations after making reference to the sketch below.
b) Show, or refer to previous results, that
Fluid A Fluid B
in which
Stainless-steel slab
(I ft X 1 ft X 2 in.)
and that for near the wall we may write Impermeable to matter and a
perfect heat insulator
pg by
v =--
z '1
...... _
............... --------------------~
0.03 ...
~ ~
N
N
::
:: where
"
,:'
z Br=
1JV2
'" k(T, - T0 )
7.8 Develop an expression for the limiting value of Nu (stagnant surroundings) for heat
transfer from a cylinder (maintained at a constant temperature) to its surroundings.
1.0
7.9 Given that the temperature distribution ofa fluid flowing within a tube is
o~~~~~2~0~0~~~---,40~0~~~~6-0~0~~~-8_JOO
T- TR [ (
To-TR= 1- R
r) 2
]
,
Temperature, F
in which TR = temperature at the wall (r = R), and T0 = temperature at the center line (r = 0);
:1:'
,,.i' also given that the flow rate may be approximated as the slug flow, that is. vz = V (constant).
I
600 X I 06 write an expression for the heat-transfer coefficient defined as follows using the mixed mean
, I
:: I
7 3.0X 106
~ temperature Tm:
~
I I
0:::
h=
400 x 106
J~ v,T2nr dr
Tm= .
J~ vz2nr dr
_,
7.10 A liquid at a temperature T0 continuously enters the bottom of a small tank, overflows
into a tube, and then flows downward as a film on the inside. At some position down the tube
(z = 0) when the flow is fully developed, the pipe heats the fluid with a uniform flux qR. The
heat loss from the liquid's surface is sufficiently small so that it may be neglected.
a) For steady-state laminar flow with constant properties develop by shell balance or show 8
by reducing an equation in Table 7.3 the pertinent differential energy equation that
applies to the falling film.
b) Write the boundary conditions for the heat flow.
CORRELATIONS .FOR HEAT TRANSFER
c) What other information must complement parts (a) and (b) in order to solve the energy WITH CONVECTION
equations?
''
The problems of heat flow with convection, discussed in the preceding chapter,
pertain to simple systems with laminar flow. The more complex nature of turbulent
flow and its limited accessibility to mathematical treatment requires, for the most
part, an empirical and experimental approach to heat transfer. This approach
leads to correlations. Please bear in mind, however, that despite the simplicity of
laminar flow problems, they should not be underestimated. Many simple solutions
Heater have been applied to real systems with approximating assumptions and, besides,
the simpler systems provide models for interpretation and presentation of complex
systems. On the other hand, the study of turbulent flow is not entirely empirical;
it is possible to establish certain theoretical bases for the analyses of turbulent
! .I transfer processes. This approach is not presented here, but other texts 1 2 provide
an introduction to this subject.
7.11 A liquid metal at the temperatures of interest has a Prandtl number of0.001. Utilize the
Figure 8.1 shows a physical picture of heat transfer in a bounded fluid. The
approximate integral technique and develop an expression for the local Nusselt number for fluid is artificially subdivided into three regions: the turbulent core, the buffer
forced convection parallel to a flat plate at a uniform surface temperature of T0 . Make the zone, and the laminar film near the surface. In the turbulent core, thermal energy
following assumptions: is transferred rapidly due to the eddy (mixing) action of turbulent flow. Conversely,
,.
a) The liquid metal approaches the leading edge with uniform temperature T00 and uniform within the laminar film, energy is transferred by conduction alone-a much slower
,'I velocity V00 process than the eddy process. In the buffer region, energy transport by both
I
.1. b) All flow is laminar. conduGtion and by eddies is appreciable. Hence, most of the total temperature
c) The velocity profile is linear. drop between the fluid and the surface is within the laminar film; across the buffer
I
d) The temperature profile is linear. zone there is an appreciable temperature drop, but within the turbulent core, the
e) The flow of heat parallel to the plate is negligible.
temperature gradients are quite shallow.
[Hint: Before you madly plunge into the problem, think about the fact that the Prandtl number
In Chapter 3, it was convenient to define a friction factor to deal with momen-
is 0.001. Also, for a linear velocity profile fJ/x = 3.46~.]
tum transport in fluids in contact with surfaces. Similarly, for energy transport
between fluids and surfaces, it is convenient to define a heat-transfer coefficient by
-k(oT/oy) 0
(8.1)
To - Tl
1
W. M. Rohsenow and H. Y. Choi, Heat, Mass, and Momentum Transfer, Prentice-Hall,
Englewood Cliffs, New Jersey, 1961.
2
R. B. Bird, W. E. Stewart, and E. N. Lightfoot, Transport Phenomena, Wiley, New York, 1960.
I
241
--------------
243
(c)
8.1.1 Dimensional analysis-Buckingham's pi theorem
' We may obtain pertinent dimensionless groups without any reference to the basic
differential equations at all if we can identify the variables sufficient to specify a
given situation. A systematic way of determining the groups is known as Bucking-
ham's pi theorem. 3 Simply stated, it reads: The functional relationship among q
quantities, whose units may be given in terms of u fundamental units, may be written
as a function of q - u dimensionless groups (the n's).
In the application of the pi theorem, these rules are helpful:
Temporal mean a) Compile a list of the q significant quantities and their respective fundamental
temperature
units. The fundamental units are L (length), M (mass), E> (time), and T
(tern pera ture ).
Cooled wall at b) Select what may be called the primary quantities or primary q's. The number
temperature T 0 ~ of primary q's should equal u, the number of fundamental units.
c) Form each n term by expressing the ratio of the remaining q's (one at a time)
Distance
from wall to the product of the primary q's, each raised to an unknown power determined
as shown below.
Fig. 8.1 Temperature profile of a flowing fluid bounded by a cooler wall; (a) laminar film, We shall now illustrate the method for heat transfer in forced convection.
(b) buffer zone, and (c) turbulent core. We presume that the heat-transfer coefficient for fully developed forced convection
in a tube is a function of the variables
where the subscript "O" refers to the respective quantities evaluated at the wall,
and TI is some temperature of the fluid. If the fluid is infinite in extent, we take TI h = f(fl, p, k, 17, CP, D). (8.2)
as the fluid temperature far removed from the surface, and designate it T00 If the
We can express all these quantities in terms of the four fundamental dimensions
fluid flows in a confined space, such as inside a tube, TI is usually the mixed mean
M, L, E>, and T. Specifically,
temperature, denoted by Tm; it is a temperature that would exist if the fluid at a
particular cross section were removed and allowed to mix adiabatically.
V, Le- 1 , 17, ML -10-1, D,L.
p, ML - 3 , C L1e-2T-1
We define the engineering units of h as Btu/hr-ft 2 F according to Eq. (8.1); as P' '
k, MLe- 3 T- 1, h, Me- 3 T- 1 ,
we shall see, his a function of the fluid's properties (k, 17, p, CP), the system geometry,
Note that heat energy is given the units of work (force x distance), namely,
the flow velocity, and the value of the defining temperature difference. We shall
ML 3 0.- 1 .
deal in the remainder of this chapter with predicting the dependence of h on these
quantities, by presenting some experimental data for various systems in the form The number of fundamental units in this case is four. Thus, since u = 4,
of correlations based on dimensional analysis. select as the four primary q's: i7, D, p, and k, leaving h, 17, and CP. Proceed to form
the first n term.
h ME>-3T-1
~ 8.1 HEAT-TRANSFER COEFFICIENTS FOR FORCED CONVECTION IN TUBES VaDbpckd = (Le-1)a(L)b(ML -3Y(MLE>-3T-1)d
In Section 3.1, we presented a method of dimensional analysis utilizing a rearrange- Equating the exponents of each dimension to zero, so that then group is dimen-
ment of differential equations in order to show that the friction factor was a sionless, we obtain
function of the Reynolds number. We used the same method in Section 7.3 for M: 1 = c + d,
natural convection. Here we present another method of dimensional analysis in L: 0 = a + b - 3c + d,
order to bring the reader's attention to a different technique of dimensional analysis, e : - 3 = - a - 3d,
while showing those dimensionless groups which are pertinent to the heat-transfer T: -1 = -d.
system of forced convection within a tube. 3
E. Buckingham, Trans. ASME 35, 262 (1915).
I
-.::::;u1re1auu11:s1u1-11t::at.-lTa11:!f1er-'ft"1u1-t..:u11l"et..:nu11~--
ir.1
z=
Similarly, the following n groups can be obtained:
nz = YffpVD
- 1
=_,
Re 0.002
-- -- .....
C~:Y/J) ~'
213
Nu = Nu (Re, Pr, L/D, Yfm/Y/ 0 ). (8.4) jH := Stm( = (8.6)
This dimensional analysis is of great use in experimental work involving heat where Stm is the Stanton number, and jH is often referred to as a j-factor. The
transfer. For example, h depends on eight physical quantities: D, V, p, Yfm, Yf 0 , CP, Stanton number is defined
k, and L. To study all combinations of eight independent variables for ten values Nu h
of each would require 108 tests, whereas, by giving Nu as a function of only four St=--=---
Re Pr CppV
groups (Re, Pr, L/D, Yfm/Yf 0 ), 104 tests would suffice. Thus, a graduate student will
We refer to Eq. (8.6) as Colburn's analogy between heat transfer and fluid friction.
have to work on a problem for only five years instead of 50,000 years!
The analogy breaks down for Rem < 10,000, and also for rough tubes because f is
8.1.2 Correlations for forced convection in tubes affected more by roughness than its counterpart j H The effect of wall roughness
i:' iI 4 was studied for air by Nunner, 5 and Fig. 8.3 relates the heat-transfer coefficient to
i: I For fully developed flow in tubes, the correlation of Seider and Tate for flow in
I smooth tubes with nearly constant wall temperature is presented in Fig. 8.2. The the ratio of the friction factor fin the rough pipe to the friction factor fs in a smooth
'.)
1,.
Reynolds number used here, Rem = DVp/Yfm, is convenient because the laminar-to- pipe of the same diameter. The results have not been tested extensively, and should
d turbulent transition is at about 2100 (the same as in Fig. 3.2), even when Y/o differs probably be restricted to use with gases.
1!
!
i
I:.'I' appreciably from Yfm. For liquid metals, where 0.005 < Pr < 0.05, the following equation represents
6
111
For highly turbulent flow (Rem > 10,000), the equation available experimental data with Re > 10,000 and for a uniform heat flux :
i1'
'~ Y/ )0.14 Nuq = 6.7 + 0.0041(Re. Pr)" 793 exp (41.8 Pr). (8.7)
i~
Num = 0.026 Re~ Pr,!( (
8 3
Y/: (8.5)
1.
5 W. Nunner, VDI-Forschungsh<ft No. 455/1956 (VDI-Verlag GmbH-Dusseldorf).
4 E. N. Seider and G. E. Tate, Ind. Eng. Chem. 28, 1429 (1936). j 6 W. M. Rohsenow and H. Y. Choi, ibid., page 193.
L
-- ::1.,
UJ Oxygen enters pipe
at80~
_ __..~
~
3 10
f/f,
Fig. 8.3 Effect of roughness on heat transfer in turbulent flow. (From Nunner, ibid.)
Solution. The oxygeq requirement, for C + !0 2 (g)-+ CO(g), is
Because of the very high values of h in liquid metals, (T0 - Tm) is not very large, so
that the mixed mean temperature properties can be used without significant error. _ (10,000)(0.20)lb C 1 lb mol C ! lb mol 0 2 32 lb 0 2 60 min 144
Equation (8.7) for liquid metals applies to uniform heat flux along the tube. Vp = (100)(5) min 12 lb C 1 lb mol C 1 lb mo! 0 2 1 hr n(t}2 ft 2
For uniform wall temperature the difference between Nur (uniform wall tempera-
lb 02
ture) and Nuq (uniform heat flux) is significant for liquid metals (Fig. 8.4); for = 235,000 hr ft2
Pr > 0.5 the difference between Nur and Nuq is negligible, and hence Eq. (8.5) is
also valid for constant heat flux boundary conditions. The oxygen enters the tube at 80F, and as it proceeds through the pipe its tempera-
ture rises. Consequently, the prpperties of the oxygen vary with distance down the
,...!<" 1.0 tube. A commonly employed estimation is to evaluate the properties of the fluid
2"... at its average temperature along the heated pipe. The heat-transfer coefficient
se, corresponding to these properties is then calculated, assumed to be constant along
0.9
the pipe, and used to determine the heat transferred to the gas.
First, arbitrarily assume that the gas leaves the pipe at 10Z0F. Then, take the
0.8 average Tm as (1020 + 80)F/2 = 550F for the last 4 ft of pipe. At 550F, the
properties of 0 2 are
1J = 0.0792 lbm/ft hr,
0.7
k = o:ono Btu/hr-ft F,
cp = 0.235 Btu/lbm F,
0.6 '-------L-----'------~
Pr= 0.690.
0.00 I 0.0 I 0.1 1.0
Pr At 2680F, 1J = 0.166 lbm/ft-hr (estimated value). Now calculate Rem:
Fig. 8.4 Comparison of uniform wall temperature and uniform heat flux in tubes. (From Rem = DVp = (1/24)(235,000) = 123 ,800 _
R. A. Seban and T. T. Sliimazaki, Trans. ASME 73, 803 (1951).)
1Jm (0.0792)
j-- Example 8.1 A 5-ton heat of steel. must be decarburized from 0.40 to 0.20% C in Thus Eq. (8.5) applies, so that
5 min. A convenient method for accomplishing this is to blow oxygen through a
submerged lance. A low carbon steel pip~ of !-in. I.D. is used as the lance, despite Num = (0.026) Re~ 8 Pr,!!
3 11~: ) 0.14
(
the fact that the end of the pipe gradually melts. If we estimate the portion of the
lance within the furnace to.be at 2680F, calculate the temperature at which the - 0.8 1/3(0.0792) 0.14
)1',, oxygen enters the liqui<;l steel. Neglect the pressure drop through the pipe, i.e., - (0.026)(123,800) (0.690) 0.166
assume that the pressure in the pipe equals 1 atm plus approximately the equivalent
= 241.
of 1 ft of liquid steel, or 1.20 atms. I
l_
Therefore
so that
]) N um = (0.0270) (241) = 156 Btu/hr-ft 2 F.
h = km 0.0360 2
1124 h = (l/ ) (243) = 210 Btu/hr-ft F,
24
An e~ergy balance applied to the fluid over a differential length of pipe, dx, can
be wntten A TL (210)(4)(n)(l/24)
In (2680 - 80) (240) (0.277)
WCP dTm = h(T0 - Tm)(nD dx),
ATL = 498F or Tm = 2182F (at x = L = 4 ft).
whe:e W = mass flow rate, lbm/hr, and T0 = temperature of pipe wall, F. Inte-
gratmg for a length of pipe L, with h, CP, and T0 taken as constants, yields We could proceed to repeat the process again using an average Tm of (2180 +
80)F/2 = 1130F. This differs by only 80F from 1050F used for the second
L !iTL
h J
0
dx = -:~P J dln(Tm - T 0), approximation. This difference would not alter the properties sufficiently to merit
further refinement.
The conclusion that the gas temperature is lower than the metal temperature
or when it exits from the lance becomes significant when the bubble size is calculated
in order to obtain the surface area and residence time for calculations of mass
In ATL = hLnD transfer. Since the gas is not fully expanded when it exits from the pipe, the bubble
A T0 size,. calculated on the basis of the pipe size, is smaller than the ultimately expanded
bubble size, and the rate of rise is faster than anticipated on the basis of the pipe
Substituting numerical values yields
size.
In ATL = (156)(4)(n)(l/24)
(2680 - 80) (240H0.235) 8.2 HEAT-TRANSFER COEFFICIENTS FOR FORCED CONVECTION PAST
SUBMERGED OBJECTS
from which
In the following correlations, the heat-transfer coefficient h is defined for the total
or Tm = 2069F (at x = L = 4 ft). surface area of the submerged object, and the defining fluid temperature is that far
Recalling our initial guess of 1020F, we see that in terms of estimating properties removed from the surface. We evaluate all the properties, however, at the film
we have not selected good values. temperature, T1 = (T0 + T00 )/2.
As a second guess, assume that the gas at the end of the pipe is at 2020F. Then, In Fig. 8.5, jH = Nu1 Re1- 1 Pr1- 113 is plotted versus Re for a long cylinder
we take the average Tm as (2020 + 80)F/2 = 1050F. At 1050F, the properties perpendicular to the fluid. The figure also shows a plot off/2 versus Re to illustrate
of 0 2 are thatjH < f /2, which is usually the case in flows with curved streamlines. Colburn's
analog:)' breaks down here because of the form drag which has no counterpart in
Yf = 0.102 lbm/hr-ft, heat transfer.
k = 0.0360 Btu/hr-ft F, These results agree closely with McAdams' correlation 7 which is based on
cp = 0.277 Btu/hr-ft F, experiments with water, oils, and air. Over the range 1 < Re1 < 10 3 , McAdams
Pr= 0.785. specifies
Re = DVp = (1/24)(235,000) _ (8.8)
m Y/m O.l02 - 96,000.
For higher Reynolds numbers, 10 3 < Re < 5 x 104 , data collected for air are
Then, using Eq. (8.5) again, we have correlated by
' \
II I:1 = 243, 7
W. H. McAdams, Heat Transmission, third edition, McGraw-HilL New York, 1954, page 268.
!1
~~-- ------ ---------------- -----------------...-...-~-----~--- ---------...-.. ~- ..-~... -oI.,..-"'""'"'"'.,-v-.'7I"~~-~UU1'~l-~-pa;:-,1.-3UITUl't;::"l"l'~UUJ~W:~------.-:JI
,~
~ 1.0 '-. 5
N ..._
v~ 4
J-1~ f'2 ::o..'
'-. 3
----i:s ~
~
~1q "' 2
1.5
II O.J
...., 10-3
"' 8
6
5
4
3
O.OJ IO' 1.5 2 3 4 5 6 8 10 6 1.5 2 3 4 5
V00 x/v1
Fig. 8.7 Heat-tra.nsfer coefficient for flow over flat plate with turbulent boundary layer and
uniform heat flux. Experimental data collected for air, and compared to the analysis for
turbulent flow (Eq. 8.12), and laminar flow (Eq. 7.27). (From R. A. Seban and D. L. Doughty,
Trans. ASME 78, 217 (1956).)
O.OOJ JO' JO' J0 4 10 5
Ref= DV=P_1/T1f Figure 8.7 shows results for the flow parallel to an isothermal semi-infinite
Fig. 8.5 Heat and momentum transfer for a cylinder perpendicularly oriented to flow. UH from flat plate. For this flow system, the Colburn analogy holds very well because there
T. K. Sherwood and R L. Pigford, Absorption and Extraction, second edition, McGraw-Hill, is no form drag; therefore, the local heat-trans.fer coefficient is related to the friction
New York, 1952, page 70;ffrom F. Eisner, see Fig. 3.9.) factor by
~Pr2/3 = fx. (8.11)
C)' '-. 1000
',, I
. .: : ;-< ~ CppVOO 2
-
II
_...., ~
z ~ For turbulent flow (from Eq. 3.28), fx = 0.0592 (Rex)- 1 ! 5 ; then Eq. (8.11) is
(8.12)
or
Asymptote
for stagnant
fluid--
]L-.L.l.J.l.J.ll.L-L..L.LJ.LWJL......J_J_Ll.llJJ..1_--'-l--1..LJLLJ..1..-'--'-'-ULI.LU
ht = Nux = 0.0296 Re 0 8 Pr 1 13 . (8.13)
_l_
-"-.----~-- - - - - - - - - -
-------~----~-------~-------------------1---.:;7v------
_, 3.2
The temperature of the wire is determined by measuring current, voltage drop, z=
hence electrical resistance, and by knowing how the resistance varies with the 0
OJ)
.9 2.8
temperature. Thus, both the heat flux-from the wire to the fluid-and the wire
temperature are measured electrically. Given air at 60F flowing normal to the 2.4
wire (0.01 ft in diameter) at 140F, with a heat flux of 2000 Btu/hr-ft 2 , determine
the velocity of the air in ft/sec. 2.0
3
p = 0.0709 lb/ft ,
CP = 0.241Btu/lbF,
k = 0.016 Btu/hr-ft F, -0.4
YJ = 0.046 lb/hr-ft,
-0.8 ';----!.---~-~---'--_L-_,1---.L---'.---.J
-4 -2 0 2 4 r, 8 10 12 14
Nu _ hD _ (25)(0.01) _ log 10 (Grr- Pr)
15 6
J - kf - (0.016) - "'
Fig. 8.8 Heat-tr;msfer coefficients for natural convection. (From W. H. McAdams, ibid.,
pages 173-176.)
i Pr = (Y/cp) = (0.046)(0.241) = 694
J k J (0.016) O. '
Figure 8.8 shows the results of correlating experimental data for free con-
1 vection from vertical plates and horizontal 'cylinders to gases and liquids with
ReJ 60 = 0. (15.6)(0.694)- 0 3 = 67.4
26 Prandtl numbers ranging from about 0.5 to 10. The effect of the variation of
properties with temperature is included by evaluating properties at TJ = !(T0 +
Ref = 1120. T00 ). The dimensionless groups are then defined as
Thus Eq. (8.9) applies as assumed, and hL
NuL =-
Y/J) (Ref) ( 0.046 ) ( 1120) kf
Voo = (PJ D = 0.0709 0.01 = 72,600 ft/hr,
Pr= vf = CpYJJ,
or rJ.J kf
V00 = 20.2 ft/sec.
GrL =L3p}gfJJ(1;_o - Too).
Y/J
8.3 HEAT-TRANSFER COEFFICIENTS FOR NATURAL CONVECTION
The characteristic dimension Lis the plate length. For cylinders, the character-
In Chapter 7, the discussion of natural convection in laminar flow led to a dimen- istic dimension is one-half the circumference, that is, nD/2; by defining L as such,
sionless equation of the form note how close the curves for the two different correlations lie. In fact, very often
Nu = Nu (Gr, Pr). no distinction is made between these two cases, and a third case, namely, that of
vertical cylinders with characteristic dimension taken as length, is included in the
This corresponds to the results of dimensionless analysis in this chapter if the
correlations.
Reynolds number, the important group for forced convection, is replaced by the In Chapter 7, Ostrach's analysis of laminar natural convection past vertical
Grashof number, the important group for natural convection.
I
-----'---
_______ T _______ _
plates led to Eq. (7.59). This relationship holds in the laminar flow region 104 <
Gr Pr < 109 for a wide range of Prandtl numbers (0.00835 :::;; Pr :::;; 1000). For
the sake of completeness let us repeat it:
NuL 0.902 Pr 112
(7.59)
YGrd4 (0.861 + Pr) 114
In the range of 0.6 < Pr < 10, the data can be represented by a much simpler
equation (still for laminar flow):*
NuL = 0.56 (GrL Pr) 114 . (8.15)
For liquid metals, the data in the laminar range for horizontal cylinders have
been correlated by 8
Pr ) 114
Nuv = 0.53 ( 0.
952
+ Pr (Grv Pr) 1i 4. (8.16)
Solution. The rate at which latent heat of fusion evolves is equal to the rate at
which heat flows from the interface into the bulk liquid
The expression is of the same form as Eq. (7.59), and also closely corresponds
numerically if the characteristic dimension D (diameter) is replaced by L taken as
nD/2.
In t4e turbulent range of 109 < Gr Pr < 10 12, the following equation has
been proposed as possibly valid for a wide range of Pr for vertical plates 9 and where Ps = density of the solid, H 1 = latent heat of fusion, and AT = TM - T00 =
horizontal cylinders with L replaced by nD/2: amount of undt;rcooling;
NuL = 0.0246Grf/ 5 Pr 7115 (1 + 0.494Pr 2i 3)- 2i 5. (8.17) dR h
-=-AT.
We use a much simpler form of Eq. (8.17) for 0.6 < Pr < 10: dt p.H1
NuL = 0.13 (GrL Pr) 113 . (8.18) From Eq. (8.19),
For a single sphere of diameter Din a body of fluid the relationship hD - (D3p2g/3AT)1/4(1'/Cp) 1/3
k - 2 + 0.060 l'/2 k
Nuv = 2 + 0.060 Grl/ 4 Pr 113 (8.19)
agrees with available data for Grl/ 4 Pr 1t 3 < 200. In most instances, if the fluid When this is substituted into the expression for dR/dt, we obtain
is a gas, this restriction means that the equation applies to small particles. Note
2 14 13
that the relationship yields Nuv = 2 when the fluid is motionless, as in Problem
7.6 where free convection is neglected.
- [1- + 0.0504 (p gf3 AT)
k AT
-dR -_ - 2
1 (1'/Cp)
- 1] .
dt p.H1 R R11 k
Example 8.3 Write an expression to relate the growth rate of a single spherical
Students of phase transformations are invited to use this as a basis for further
nucleus of solid as it develops in a pure supercooled liquid. Express the result in discussion and/or refinement.
terms of dR/dt as a function of supercooling AT and the properties of the metal.
The thermal profile is depicted below; the temperature of the solid may be assumed
as the freezing point TM.
* To make a distinction between plates and cylinders in Fig. 8.8, the coefficient 0.56 in Eq. (8.15)
should be replaced by 0.59 or 0.53, respectively.
8
Reactor Handbook 2; Engineering, AECD-3646 (May, 1955), page 283.
9
W. M. Rohsenow and H. Y. Choi, ibid., page 204.
Fig. 8.9 The convection pattern over a horizontal surface which is warmer than the fluid.
I
In the previous discussion of this section, heat transfer took place in the presence I ntcrface evap. Bubbles Film
of convection mostly in the vertical direction. The convection behavior in the II Ill IV v VI
vicinity of horizontal surfaces is quite different, as shown in Fig. 8.9, and the
01)
correlations of Eqs. (8.15)-(8.19) do not apply. McAdams 10 recommends to use Pure convection heat -~
the following equations for some situations of natural convection from horizontal ,,,
0
transferred by 'o
.D
--' superheated liquid .
surfaces: rising to the liquid- ~
in the laminar range, 10 5 < GrL < 2 x 10 7 . 0.1 1.0 10 100 I,000 I0.000
2. For a square plate with a surface warmer than the fluid and facing downward,
or cooler surface facing upward: Fig. 8.10 Characteristic boiling curve. (From W. M. Rohsenow, Developments in Heat
NuL = 0.27(GrL Pr) 114
. (8.22) 'I'ransfer, MIT Press, Cambridge, Massachusetts, 1964, Chapter 8.)
In Eqs. (8.20)-(8.22), L is the length of the side of the square. As approxima- conducted considerable research into the boiling process, which has resulted in a
tions, we can also apply the equations to horizontal circular disks with L defined number of correlations for boiling heat transfer.
as 0.9 D (D being the disk diameter). Although boiling is a familiar phenomenon, the next section shows that from
the point of view of energy transport, it is a complicated process. Figure 8.10
8.4 QUENCHING HEAT-TRANSFER COEFFICIENTS illustrates the complexities of boiling in which several regimes exist. If a metal
surface capable of being heated is submerged in a pool of liquid at its saturation
In the next chapter, we present analytical expressions which describe the tempera- temperature, T.a., the following events take place. As the surface temperature Tw
ture during heating or cooling as a function of time and position within a solid. is raised above T. ., convection currents circulate the superheated liquid. Heat
The most difficult variable to ascribe to such unsteady-state situations is probably transfer with convection takes place, and the correlations discussed for natural
the heat-transfer coefficient governing the energy transport between the solid's convection in Section 8.3 apply. However, the liquid cannot be indefinitely
surface and the surroundings. The methods of estimating h for some convection superheated so that further increases of (Tw - T.at) above approximately 4F are
systems, which were presented in previous sections of this chapter, can be applied accompanied by vapor bubbles forming at preferred sites on the surface. This is
satisfactorily to certain problems of unsteady-state heat flow. For the most part, regime iI in which most of the bubbles do not exceed a certain size necessary
however, correlations of h-values applicable to quenching operations have not for their escape. In regime III, (Tw - T.at) is large enough so that larger and more
been obtained because of the complexity of the convection systems. stable bubbles grow, and at the same time more bubbles form because more
Despite the importance of quenching operations in the heat treatment of metal nucleation sites become active gn the solid's surface. This mechanism, which is
parts, metallurgists have not considered nearly enough of the quantitative aspects called nucleate boiling, can generate very high heat fluxes. As the bubbles rapidly
of quenching heat transfer. The most quantitative work related to quenching has form and detach themselves from the surface, fresh liquid rushes to the former
actually been developed in the field of boiling heat transfer. Historically, the bubble site, and the very rapid process of bubble formation and detachment repeats
primary purpose of boiling heat transfer has been simply the conversion of liquid itself over and over again. The net result is that all the bubble sites act as micro-
t I into vapor, and perhaps this is the reason why this problem has not been studied pumps and create a large degree of convection at the surface.
'
enough by metallurgical engineers. On the other hand, engineers, who are involved If (Tw - T.at) is raised to even higher values, for example, 70F for water, then
with the development of nuclear reactors, rocket nozzles, and spacecraft, have the nucleation sites on the surface become so numerous that the bubbles actually
coalesce, and tend to form a vapor blanketing layer. Beyond this point, sometimes
1
0 W. H. McAdams, ibid., page 180. referred to as the burnout temperature, the heat flux decreases with increasing
--..--- --------- -zss----COrrelations for lieanransfer witli convection---- ----- ------------------------ - ---------------is;ii.-----...--~-- 0 ------------ ----- QUen-cnmg-n-e-at=trau-srer coernc1ems-I59
(Tw - T..,) because of the vapor's insulating nature. Regime IV encompasses this
rI
Table 8.1 Severity of quench values and heat-transfer coefficients for steel*
effect wherein an unstable vap6r film forms but collapses and re-forms rapidly.
For values of (Tw - T..,) greate~- than about 400F (for water), a stable film exists \ Quench media H-value, ft- 1 h, Btu/hr-ft 2 F
(regime V). This is called.film boiling, with which rather low heat fluxes are asso-
Oil, no agitation 2.4 48
ciated (contrasted to nucleate boiling); for values of (Tw - T..,) greater than Oil, moderate agitation 4.2 84
1200F, radiation across the vapor blanket becomes important, and the rate of Oil, good agitation 6:0 120
heat transfer increases accordingly. Oil, violent agitation 8.4 170
;- 1000-
Water, no agitation 12.0 240
,, Water, strong agitation 18.0 360
~
~,
200
In addition, during quenching h varies with the temperature of the solid itself.
O~-~-~-~-~~
The values in Table 8.1 are derived specifically to predict the time to cool steel from
0 I0 20 30 40 50 its austenizing temperature to a temperature midway between the austenizing and
Time, sec
original quench temperatures. Therefore all that we can say about the above
Fig. 8.11 Cooling curve of a metal part quenched in water. h-values is that they are average and apply during cooling from 1500F to about
780F. It is left to the reader to comprehend the "political" terms-moderate
Returning to quenching, we see that Fig. 8.11 shows a typical cooling curve of agitation, good agitation, and violent agitation-given in Table 8.1.
a metal part plunged into water. Note the three distinct stages during the cooling For precipitation-hardenable aluminum alloys, an important step in their heat
process. Stage A, called the vapor cooling stage, occurs immediately upon im- treatment is the cooling rate during quenching from the homogenization tempera-
mersing the part in the water. Liquid vaporizes adjacent to the surface and forms a ture. If the cooling rate is too low through a critical range of about 800-500F,
continuous vapor film. Cooling during this stage proceeds by film boiling and, if then precipitation occurs, and subsequent ageing cannot be controlled. On the
the part is at a sufficiently high temperature (i.e., steel), also by radiation. The other hand, if the cooling rate during quenching is too great, then untolerable
second stage, B, is often denoted by metallurgists as the "vapor transport stage." distortion in many parts occurs. In practice, therefore, heat treaters usually quench
This term is not really appropriate because during this stage there is a transition aluminum alloys in heated water to slow cooling and minimize distortion. In
from partial nucleate boiling and unstable film boiling to entirely nucleate boiling. recent years, some heat treaters have substituted water-polyalkylene glycol (PAG)
During stage C, referred to as the liquid cooling stage, there is no vapor formation, solutions for heated water. These solutions lower the cooling rates, and may be
and cooling occurs with natural convection. Thus we see that as the metal cools substituted for heated quench water; in addition, more uniform cooling is attained,
from its initial elevated temperature to the fluid's temperature, it passes through all and distortion can be reduced.
the regimes of the boiling curve (Fig. 8.10). Figure 8.12 gives some values of h for water-PAG solutions based on experi-
For these and other reasons no comprehensive body of engineering correla- mentally measured cooling rates through the range 800-500F. When solutions
tions has evolved and been applied to quenching problems. Rather most of the of PAG and water are used, his dependent on composition. In these instances, an
literature gives some average values of h which apply to specific cases. immiscible PAG-rich liquid phase forms at about 165F (the exact temperature
For example, we can find the vaiues of a parameter H, known as the severity depends on the composition). Thus, near the hot solid, the liquid separates into
of quench, and defined as h/k, which represent quenching practices for steel. two immiscible liquid phases, and the surface becomes coated with the very viscous
Table 8.1 gives some typical values of H. PAG-rich phase. Therefore, the rate of heat transfer is not controlled by a water
Data from this table should be applied only to conditions for which the values vapor film whose thickness is dependent on bulk water temperature to a large
are derived. For example, h can depend strongly on the temperature of the liquid extent, but rather by the layer of the PAG-rich phase whose thickness and proper-
media alone. The values shown above are for quenchants at room temperature. ties are presumably more predictable and not so much dependent on the bulk
!;'-- 4
2
/
......... "'''\
I .
Oil designation /,'"- \
----Slow
- - Intermediate
./ \
.{J \I
\\
- - Fast /J .
4'
~
\l \\. ________ _
..d~
~--
--- . ~._,,,,.~
- ---- --
10 ~~2~00~~---;:40~0,--~-6~o~o~L.._~s~oo,,_~~1=0~00=--~~1~20~0~.l..._-14~0-o--'-1-6_.oo
"'"'
8000 4 x 10 6 N
4000
!-"
0 "'"'
0
N N
.;
.<: ;:;- ~
;:;- ~ -"
;:;-
~ ~
,., >< .,,~.
-
::l
6000 3 <;:::
""'
;:;
E :r:"'"' "
"' 3000
;:;
<;;;:
'-
"'u0 0"'
...u
~ ~
~ 4000 2
t:: ""'
q~ )
I
't
:!
:r:"'"'
~
:r:"'"' 2000
l 2000 IX 106
J
fi 0 0
1000
Fig. 8.15 Heat-flux and heat-transfer coefficients for cooling a horizontal surface from above
with a fan-type spray of water at 70F. (Adapted from P. M. Auman, D. K. Griffiths, and o';:"-------=-=-=-------:~----:-::'=-=----~
0 400 800 1200 1600
D.R. Hill, Iron and Steel Engineer, Sept., 1967.)
Surface temperature of a solid, F
Fig. 8.16 Heat-transfer coefficients for oil-spray quenching and immersion in oil. (a) Very
increase in the heat flux at about 1200-1000F, which represents a transition from strong spray; flow rate >17.2gal/sec-ft2. (b) Strong spray; flow rate= 14.7gal/sec-ft 2 .
boiling with a vapor film to nucleate boiling. (c) Still oil. (From Zimin, ibid.)
When the vapor film is present during spray cooling, the rate of heat transfer
depends on the degree to which the steam film can be broken down by the impinging found in a study, 13 that the cooling rate due to properly applied oil sprays was
droplets. With sprays, the droplets approaching the hot metal surface encounter three to five times the cooling rate by immersing in oil in the conventional manner.
the vapor film, and their success in penetrating this film depends on their kinetic Figure 8.16 shows some of these data.
energy. If the sprays are placed closer to the metal surface or the water velocity is
increased, thereby increasing the droplets' kinetic energy, then the probability that 8.5 BOILING HEAT TRANSFER
the droplets will penetrate the film increases and so does the rate of heat removal.
With this in mind, it has been a(gued that a continuous stream of water, In the past, metallurgists have made hardly any effort to apply boiling heat-transfer
because of its high kinetic energy associated with its large mass, should penetrate theory and correlations to quenching processes. However, they should be aware
the vapor film and bring about an increase in heat-transfer rates. 12 For this reason, of this field of technical literature because many of the correlations could be
large nozzles in combination with low water pressure have been utilized to produce directly applied to quenching, or serve as a basis for a better understanding of the
a falling stream, or jet, of water which does not break into droplets. For such a jet, role of properties of fluids used as quenchants.
an increase in the flow rate of water increases the rate of heat transfer up to a point. Because boiling heat transfer is so complex, it cannot be expected that a single
If the jet-flow rate passes a critical value, then the falling jet breaks up into less equation could possibly correlate data over all the regimes shown in Fig. 8.10. For
.'',.
effective droplets, and the rate of heat transfer decreases. this reason we present, under separate headings, some available correlations
In addition to water sprays, oil sprays have been used for quenching. It was applying to individual boiling regimes.
13
N. V. Zimin, UDC 621.784.06, pages 854-858, Plenum Press, New York, 1968. Translated
12 from Metallovedenie i Termicheskaya Obrabotka Metal/av, No. 11, 62-68 (Nov., 1967).
E. R. Morgan, T. E. Dancy, and M. Korchynsky, J. of Metals 17, No. 8, 829-831 (1965).
-,..,,,.
nonmg-ii..:arrran~n:er-.
8.5.1 Film boiling We use Eq. (8.26) to determine the velocity profile by subjecting it to the usual case
of no slip at the wall, that is,
An equation often referred to for calculating the heat-transfer coefficient in film
boiling is that of Bromley 14 which applies to horizontal cylinders and vertical B.C. 2 V2 = 0 at}'= 0. (8.27)
plates submerged in liquids at their saturation temperature. The momentum and When this is done, we obtain an expression for the velocity profile:
energy transport involved in the continuous vapor film is a good deal simpler
than the transport in the other boiling regimes. For this reason, a simple analytical v = g(p, _. p)[c)y - y2] (8.28)
model has been found as a good basis for analyzing data. z 11 2
from Eq. (8.28), we determine the average velocity of the vapor
Vapor
j7 =~
c)
f
d
Dz
d
}'
= g(p, - p) c)Z
311 .
(8.29)
0
We can write a heat balance over the vapor film thickness c5 for a length dz.
For illustrative purposes we assume a linear temperature profile across the vapor
film. The heat flux across the film is (k/c5)(Tw - T.at), and this energy goes into the
Liquid
enthalpy change of the liquid in vaporizing to gas and superheating to the average
vapor temperature. Thus
d
Fig. 8.17 Vapor film model. Here dw represents the mass of vapor formed in the element dz, and Rv is the heat
of vaporization. When we substitute Eqs. (8.28) and (8.29) into Eq. (8.30), and carry
Consider the stable film of vapor in laminar flow along the vertical wall in out the integration, the result is
Fig. 8.17. For a steady incompressible vapor flow, a momentum balance applied
to the unit element yields
~c) (Tw - T )d
sat Z
= Rv [1 + ~ 8 Cp(TwR-
v
T.at)J d W. (8.31)
dryz dP (8.23)
-d =-d +pg. The entire coefficient of dw on the right-hand side ofEq. (8.31) is sometimes referred
y z
to as the effective heat of vaporization, R~.
Since this vapor film is parallel to a liquid column, we can write the pressure term A more detailed analysis of the energy transport gives a somewhat different
as dP/dz = -gp 1, where the subscript l here and within the remainder of this value of H~ than that above and should be utilized for calculations. Bromley
argument refers to liquid properties. With this substitution, Eq. (8.23) becomes uses the following definition for the effective heat of vaporization :
dryz/dy = g(p - pi). (8.24)
We solve Eq. (8.24) subject to the boundary condition that the liquid moves freely
H' = R
v v
[i + 0.4Cp(Tw -
RV
T.at)]z (8.32)
and the average coefficient is given by The discussion thus far-and accordingly Eq. (8.38)-applies to film boiling
L in liquids at saturation temperature. Film boiling in a subcoo/ed liquid is of greater
hL = J hz dz,
interest to metallurgists. This situation has been analytically treated by Sparrow
and Cess, 15 who solved the continuity, momentum, and energy equations for both
0 the vapor and liquid phases in film boiling next to a vertical plate with no forced
finally yielding 0 convection. The final result of their analysis was closely fitted by the equation
1
------------------~------------------ ..------- -
----------uu111ni;-ucn1,-r1""D1~ci-..:;~------
R = 0.05 R = 0.10
(f'Y.i 1.2 e Emery 320
"- Emery 60
O LapE
Mirror finish
0.8
0.30
0.6
0.50 0.4
0.2
0 1.4 1.6 1.8 2.0 0 0.2 0.4 0.6 0.8 I .0 1.2 1.4 I .6 1.8 2.0
(a) R = 0.05 CJ,6TIH,
C,,6T/H,.
" (b)R=0.10
R = 0.15
R = 0.25
f'Y.i 1.2 irv. 1.2
1.0
1.0
0.8 0.8
0.6 0.6 1000
6 T, F
- PrL = 2 - PrL =2 Fig. 8.19 Effect of roughness on boiling heat transfer; test results are for copper-pentane.
0.2 --- PrL =I 0.2
--- PrL =I (From P. Berenson, J. Heat Transfer, ASME, Aug. 1961, page 35.)
! . . . 1 .. I ,!, I I' .. 1,,,1
0 0.2 0.4 0.6 0.8 I .0 1.2 1.4 I .6 1.8 2.0 0 0.2 0.4 0.6 0.8 I .0 1.2 1.4 1.6 I .8 2.0 then, finally, the result obtained by Sparrow and Cess reduces to
Cp6T/H,
(c)R = 0.15 (d) R = 0.25
[ 1JD(Tw -
H~gp(p, _ p)k3] 114
Fig. 8.18 The computational parameter f!J for film boiling calculations. (From Sparrow and hn = 0.60 . (8.47)
Cess, ibid.) T.at)
This equation agrees almost exactly with Bromley's result-Eq. (8.38)-except
can be tested at two extremes, specifically for the case of no subcooling and for the that there is a slight difference in the values of H~, Eq. (8.32) versus Eq. (8.46).
case of large subcooling. In the former case, T.at - T00 = 0 and f!J = CP(Tw - However, in most cases of practical interest Cp(Tw - T.a 1 )/Hv < 0.5, and the two
T.a 1 )/Hv. Substituting this value of f!J into Eq. (8.41), we write an expression for are almost exactly equal._
film boiling with no subcooling: Sparrow and Cess examine their result for large subcooling. In Fig. 8.18,
~ertical markers appear on the curves; these markers indicate loci, the left of which
Nu 1[
16v2p J 1/4 4[
= - 0.84 +
Hv J 1/4
(8.44)
actually give results that are exactly equal to the Nusselt-Grashofrelationship for
g(p, - p)L 3 3 Cp(Tw - T.atl purely free convectional heat transfer next to a vertical surface. Thus, for physical
conditions lying to the left of the vertical markers, the problem is actually one of
Ifwe now assume that this resvlt for vertical plates oflength Lis valid for horizontal natural convection, and Eq. (8.41) need not be used.
cylinders with L taken as nD/2, then we can rearrange Eq. (8.44) (remembering that
Pr ~ 1, and hence k/17 ~ CP) as 8.5.2 Nucleate boiling
In most industrial, but nonquenching, boiling applications, nucleate boiling occurs
more frequently than film boiling. Therefore the literature on nucleate boiling is
(8.45)
more voluminous than that on film boiling, and there are many correlations
[!
available that attempt to predict the rates of nucleate boiling heat transfer. How-
I I If we define the effective heat of vaporization as ever, the correlations often disagree with one another, probably because nucleate
boiling is really an ill-defined process.
H~ = Hv[ 1 + 0.84 Cp(TwH~ I'.at)J (8.46) Since nucleate boiling involves the nucleation and growth of vapor bubbles
that originate at nucleation sites on the metal surface, such as cavities, scratches,
Table 8.2 Values of Csf for Eq. (8.51)*
Water-nickel 0.006
Water-platinum 0.013
Water-copper 0.013
Water-brass 0.006
log (Tw-Ts) CC1 4 -copper 0.013
Benzene-chromium 0.010
Fig. 8.20 Effect of subcooling on boiling heat transfer.
n-pentane-chromium 0.015
Ethyl alcohol-chromium 0.0027
." '' etc., it is rather complex. Some of the variables which affect the rate of heat Isopropyl alcohol-copper 0.0025
transfer are : 35 % K 2 C0 3 -copper 0.0054
1. Surface condition. Changing the surface by polishing causes the q versus 50 % K 2 C0 3 -copper 0.0027
(Tw - Tsai) curve to shift toward the right, as shown in Fig. 8.19. Surfaces that n-butyl alcohol-copper 0.0030
have not been freshly machined or polished often require a higher (Tw - 'f..1) for
the same q in the nucleate boiling regimes. This may be due to a decrease in cavity * From Rohsenow, ibid.
sizes resulting from mild oxidation, thus requiring more superheat for equivalent
amount of bubble formation. We omit the details of Rohsenow's analysis here, but the final equation, based
2. Gases. Gases dissolved in the liquid, which come out of solution at the surface on the above dimensionless groups, which correlates experimental data, is
well below the normal saturation temperature, agitate the liquid in the same way
as vapor bubbles; thus, data show higher heat-transfer rates than those obtained
, Cpz(Tw - 'f..,) = C.1[_!!_ ~] 1/3 Prz17. (8.51)
Hv 1'/zHvy~
for a degassed liquid.
One correlation for nucleate boiling relies on the supposition that the heat- Here (J is the liquid-vapor surface tension, and c.1 is a constant dependent on the
transfer rate in boiling results from the increased convection of the liquid. With nature of the surfaces and the fluid f. Values of c.1 deduced from the experimental
this in mind, Rohsenow 16 proceeds to correlate data by similarity with single-phase data are given in Table 8.2. This correlation applies when the bulk liquid is at
forced convection heat-transfer, assuming that saturation temperature, and is also valid ifthe liquid is subcooled. The subcooling
does, however, shift the characteristic boiling curve in the nucleate boiling regime
Nub = f(Reb, Pri). (8.48)
(Fig. 8.20).
The dimensionless groups are defined as
S.5.3 !ransition boiling
N _ qwD,, hDb By using the results of Sections 8.5.l and 8.5.2, we can describe the heat flux for
(8.49)
ub = (Tw - T..,)kz = k;' film boiling and nucleate boiling. The transition between these two, depicted as
and regime IV in Fig. 8.10, may be described by evaluating the critical heat flux qmax at
the burnout temperature, and the minimum heat flux qmin corresponding to the
GbDb beginning of stable film boiling. Values of the heat flux in the transition regime lie
Reb =-- (8.50)
111 on the straight line joining qmax and qmin on a log q versus log AT plot (Fig. 8.19).
The magnitude of qmax seems to be relatively insensitive to the nature of the
Here Db is the diameter of the bubbles as they leave the heating surface, and Gb is
surface finish, because the controlling physical phenomena in the boiling now take
the rate of mass of vapor (or bubbles) formed per unit area of heating surface.
place at the vapor-liquid interface removed from the solid's surface. One descrip-
tion of what happens at the critical heat flux is attributed to Zuber 1 7 who used a
17
16
W. M. Rohsenow, Developments in Heat Transfer, Chapter 8, MIT, Cambridge, Massachu- N. Zuber, Hydrodynamic Aspects of Boiling Heat Transfer, U.S. Atomic Energy Commission,
setts, 1964. AECU-4439, June, 1959.
I
purely hydrodynamic argument. As the critical heat flux is approached, the Solution
amount of vapor entering the liquid as undulating columns increases, and thus a) For water at 212F, apply either Eq. (8.38) or (8.47). Using Eq. (8.38), one
reduces the available area for liquid flow toward the surface. Therefore, the obtains
relative velocity between the liquid and vapor increases to the point where the
vapor-liquid interface becomes unstable and prevents the liquid from contacting
the heating surface.
Zuber's hydrodynamic theory leads to the following expression for the
maximum heat flux in pool boiling of saturated liquids:
qmax = 0.18[grr(p, 2- p)J 1/4[_Pz_J 112. (8.52) ("Steam Tables" are a convenient source for many properties of water and steam.)
pHV p P1 + p At 1 atm and 212F
For boiling of subcooled liquids, the problem is to determine the energy Hv = 970.3 Btu/lb, p 1 = 59.8 lb/ft 3 .
transfer from the interface into the bulk liquid. In accordance with the hydro-
dynamic aspect of the problem, Zuber reasons that because the vapor is periodically For water vapor at 1 atm and an average film temperature of (1500 + 212)F/2 =
released from the vapor-liquid interface, the temperature distribution within the 856F, the following properties apply
liquid is also periodically renewed by conduction of energy into the newly arrived
- subcooled liquid at the interface. The added energy term results in this expression
cp = 0.502 Btu/lb F,
p = 0.0188 lb/ft 3 ,
for boiling of subcooled liquids at bulk temperature I;,:
k = 0.034 Btu/hr-ft F,
112 [g(pz _ p)J 1/4[rr(p, _ p)g] 1;8 '1 = 0.062 lb/hr-ft,
qmax-sub = qmax + 0.696(T.at - J;,}(kzpzCpl} (J p2
2
H' = 9703 [ 1 (0.4)(0.502)(1500-212)] ,
(8.53) v + 970.3
For the minimum heat flux, Zuber's hydrodynamic theory leads to an expres- = 1558 Btu/lb.
sion which has wide acceptance :
Then
grr(p, _ p)J 1/4
qmin = 0.128 [ (pz + p)2 (8.54) 2 3 114
h = 0.62[(1558)(32.2)(3600) (0.0188)(59.8 - 0.02)(0.034) ]
(1500 - 212)(0.062)(1/24)
The constant 0.128 is recommended by Kesselring et al. 18
D
For boiling of subcooled liquids, if the added energy conduction into the liquid 2
= 171 Btu/hr-ft F.
is determined for the hydrodynamic conditions corresponding to qmin then one
Now determine the effect of radiation with i:;w = 0.78.
can develop the following equation:
4
h = i:;wrr[Tw - T.!1J
112 [g(p, _ p)J 1/4[rr(p, _ p)g] 1;8
qmin-sub = qmin + 0.775(T.at - Tb}(kzpzCpl) (J pf r Tw - T.at
(0.78)(0.1713)[19.6 4 - 6.72 4 ]
(8.55)
(1500 - 212)
Equation (8.55) has not been tested experimentally, but is recommended for use
pending verification. = 15.3 Btu/hr-ft 2 F.
Example 8.4 Estimate t:1e initial heat-transfer coefficient that applies when a Then we find the total heat-transfer coefficient by using Eq. (8.39):
!-in. diam. steel bar at 1500F is quenched in water at (a) 212F, and (b) at 70F.
The bar is held in the horizontal position. h= hv(h:) + h, = 171(1~ ) + 15.3 1
18
R. C. Kesselring, P.H. Rosche, and S. G. Bankoff, AIChE J. 13, 669-675 (July, 1967). = 179 Btu/hr-ft 2 F.
------------------------~P~r~ob~re_m_s-~27~5~----
----- ----------~--------- -- ........--.._.......""'"'""""".
....
b) For water at 70F, apply Eq. (8.41). First R, as given by Eq. (8.42), must be
calculated with vapor properties evaluated at 856F as before and liquid properties
at (212 + 70)F/2 = 141F. At 141F, the properties for liquid water are
P1 = 61.4lbm/ft3,
111 = 1.21 lbm/hr-ft,
/31 = 0.29 x 10- 30 R- 1 ,
CP 1 = 1.00 Btu/lbmF,
Pr1 = 2.96,
k 1 = 0.409 Btu/hr-ft F.
2000F flows with an average velocity-of 50,000 ft/hr. The environment surrounding the tube
112 114 4
(0.0188)(0.062)] [61.4 - 0.02] [ i.oo 3 T1 is air at 100F, and the outside surface temperature is 150F.
R = [ (61.4)(1.21) 0.0188 (0.29 x 10- )(970.31] ,
Data:
= 0.129.
Material T,F p, lbm/ft
3 I], lbm/hr-ft k, Btu/hr-ft F CP, Btu/lbm-F g/J/v2, (ft3_oFr i
cp 1 fi'I'i (1.00)(142)
= 0.146,
= (970.3) 3
Hv Gas 2000 0.02 0.1 * 0.05 0.25
Air 100 O.D7 0.049 0.016 0.25 2
CP!iT = (0.502)(1288) = 0.666. Air 150 0.06 0.053 O.Q18 0.25
Hv 970.3
*At these high temperatures, IJ for the gas does not change much with temperature.
For R = 0.10, Fig. 8.18b yields r!J 1i 4 2:'. 0.81; for R = 0.15, Fig. 8.18c yields
8.2 A vertical surface 5 ft high is at 150F and the ambient air temperature is 50F.
r!J 1 i 4 2:'. 0.85. Therefore,atR = 0.129,& 114 2:'. 0.83,andf!JJ = 0.954. Nowsubstitute
values in Eq. (8.41) with L = nD/2. a) Calculate the heat-transfer coefficient using information given in the text.
b) Repeat using one of the following simplified equations, which apply reasonably well to
2 114
Nu = [0.
84
l .O
5
J114 [(32.2)(3600) (61.4 - 0.02)(n/2)(1/24)] air, CO, N 2 , and 0 2 in the range 100-1500F. Land D are in ft, !J.T in F, and h
I 3 + (16)(3.30)2(0.0188) in Btu/hr-ft2 F.
l
8.4 Air flows at 10 ft/sec through a round tube (i-in. diameter by 12 ft long) with a uniform
tube wall temperature of200F. If the air enters at 60F, at what temperature does it exit from
the tube?
8.5 Oil flows in a long horizontal 2-in. I.D. copper tube at an average velocity of 10 ft/sec.
9
If the oil has a bulk temperature of 200F and the air surrounding the tube is at 70F, calculate:
a) the "liquid-side" heat-transfer-coefficient; CONDUCTION OF HEAT IN SOLIDS
b) the "vapor-side" heat-transfer-coefficient;
c) the temperature of the copper tube;
d) the rate of heat transfer to the air.
~8.6 Replace the oil with sodium and repeat Problem 8.5 (a)--(d). Compare or contrast the
results of the two problems.
Practicing metallurgists easily recognize that the conduction of heat within solids
8.7 Water flows through a tube 3 ft long by 1-in. I.D. at a velocity of 15 ft/sec. The tube wall
is fundamental to understanding and controlling many processes. We could cite
temperature is kept constant at 210F by condensing steam. If the inlet temperature is 60F,
numerous examples to emphasize the importance of this topic. Some important
calculate the exit temperature.
metallurgical applications that fall in this category include estimating heat losses
8.8 A heat-treating furnace is 20 ft long, 10 ft wide and 6 ft high. If a check with thermocouples from process equipment, quenching, or cooling operations where the cooling rate
indicates that the average wall temperature is 150F and the top is at 200F, calculate the heat of a part actually controls its metallurgical state and hence its application, and the
loss from the furnace in Btu/hr. solidification of ingots or castings.
8.9 In flow past a flat plate, a laminar boundary layer exists over the forward portion between
O and L,,, and the turbulent boundary layer exists beyond L,,. With this model, the average h
9.1 THE ENERGY EQUATION FOR CONDUCTION
I: over a plate of length L (with L > L,,) can be determined as indicated
f. Ltr L The general equation for the conduction of heat in a solid free of heat sources or
':fI
ii,,,
h= ( f hx(lam) dx + f hx(turb) dx) .
sinks can be written
0
' Ltr
(9.1)
I Take Re,, = 3.2 x 10 5 and show that
The more common form of this equation is written for conductivity independent
l hL/k = 0.037 Pr 113 (Re 0 8 - 15,500). of position in space and specific heat independent of temperature:
[Hint: L,,/L = Re,,/RevJ 2 2 2
aT (a T a T a T)
\
8.10 Molten aluminum is to be preheated while being transferred from a melting furnace to a
at = a v2 T = a ox2 + oy2 + oz2 . (9.2)
casting tundish by pumping it through a heated tube, 2 in. in diamceter, at a flow rate of 10,000
;i lb/hr. The tube wall is kept at a constant temperature of 1400F. where the thermal diffusivity, as defined in Chapter 7, is
1:, a) Calculate the heat-transfer coefficient between the wall and the aluminum. a= kjpCP. (9.3)
b) Using this value of the heat-transfer coefficient, how long would the tube have to be to
Equation (9.2), although not as rigorous as Eq. (9.1), is used when analytical
heat the aluminum from 1250 to 1395F?
'i
.jlr'.
~
"! ".
~ :
'
Data for aluminum at 1300F:
k = 50 Btu/hr-ft F,
solutions are sought.
When the temperature in a body is not a function of time but only depends
upon position in space, then Eq. (9.2) becomes
p = 160 lb.,jft 3 , a2T a2T a2T
Ii V2T = ox2 + oy2 + oz2 = 0. (9.4)
. ' CP = 0.25 Btu/lbm-F,
'1 = 2.9 lb.,jhr-ft. Equation (9.4) therefore applies to steady-state conduction in systems free of heat
sources and sinks. It is often referred to as the Laplace equation.
277
1
-:r__ _
l,
t T;
i T,
To
k1 k2
/' ~J,
i I x =L 2
I
x = 0 x=L
Fig. 9.1 Steady-state temperature distribution in a plate. Fig. 9.2 Steady-state temperature distribution in a composite wall.
9.2 STEADY-STATE ONE-DIMENSIONAL SYSTEMS We can build up a solution based on the four equalities in Eq. (9.10). However,
this procedure is rather tedious, and it is much easier to make use of the resistance
9.2.1 Infinite flat plate concept.
Equation (9.4) for an infinite flat plate, such as that in Fig. 9.1, reduces to The flow of heat Q through material, subject to a temperature difference
d 2 T/dx 2 = 0. (9.5) ~ - ~, is analogous to the flow of current I, as a result of a potential difference
Boundary conditions are Ei - Ek through an electrical conductor. From Ohm's law for electricity, the
thermal resistance R 1 for heat flow is visualized:
B.C. 1 at x = 0, T = T1 ; (9.6)
J__
r
----------------------------------------- .,------------;,~--""--I-----
outside surfaces are 5 and 1 Btu/hr-ft 2 F, respectively. The gas temperature inside All other temperatures are determined in the same manner:
the furnace is 2000F and the outside air temperature is 90F. T1 = 1952F,
a) Calculate the heat-transfer rate through the wall (Btu/hr-ft 2 ). T2 = 1649F,
b) Determine the temperatures at all interfaces. T3 = 1346,F,
T4 = 337F,
T5 _= 332F.
T, = 2000F--- In addition to the resistances discussed in this section, composite walls often
h; = 5
have another type of thermal resistance. When two elements of the composite are
I
I in contact, a resistance occurs which depends on the roughness of the two surfaces,
the fluid between the surfaces, and the contact pressure. For furnace walls, it is
customary to ignore the additional resistance because of the very high thermal
resistance of the refractories; for metallic composite walls, however, it is wise to
include it. Unfortunately, reliable data are not available because almost each
individual situation is different, and so experimental information must be obtained.
T0 = 90F . An example of a situation where the additional resistance between layers has
l:!o =I been determined can be found in the field of steel coil annealing. Olmstead 1
k = 0.60 k = 0.40
k=
0.04 k ='261 111 0 ' / determined that the effective thermal conductivity of the gap between layers of
~\' metal in a coil is 0.0446 Btu/ft-hr F. Even though ksteet = 26.0 Btu/ft-hr F, the
i---1 -9in_-'---i----i6 .1.J ~Ln. average thermal conductivity in a radial direction (perpendicular to the sheet) is
only 1.79 Btu/ft-hr F. The result of this study was a redesign of coil heating
systems to emphasize heat transfer to the ends of the coil rather than the sides.
11 - \o 1-
Solut~onQ.
C2/ - __.;.. . '&
-ii - \ . . , /A- 1.~." \.
_,_"""
'I..
,.-;s,:. 1(v 9.2.3 Infinite cylinder
A I; - T0 i (2000 - 90) For steady radial flow of heat through the wall of a hollow cylinder as depicted in
a) q = . AR 1/ / (~) (9/12) (6/12) (2/12) (1/96) ~.
Fig. 9.3, the Laplace equation still applies. However, in this situation, cylindrical
coordinates are convenient. The Laplace equation written in cylindrical coordi-
- 5 + 0.60 + 0.40 + 0.04 + 26 + 1 nates can be deduced from Eq. (B), Table 7.5.
2
= 242 Btu/hr-ft .
1 a{ oT) 1 aT aT
2 2
(9.13)
b)
'
q = hi(I; - Ti). -;: or ra;: + r2 ae + oz = O.
2 2
Therefore For the case at hand, temperature depends only on the radial coordinate;
therefore, Eq. (9.13) reduces to:
242
I; - Ti = q/h; = S = 48.4F, ~ ~(r dT) = 0. (9.14)
r dr dr
T1 = 1952F.
The boundary conditions under consideration are
Similarly,
"~ B.C. I at r = r 1 T = T1 ; (9.15)
T _ T = ~ = (242)(9/12) = 0
B.C. 2 at r = r2 T = T2 (9.16)
1 2 (kif Li) . 0. 60 302.6 F,
and 1
C. F. Olmstead. Theory and t(volution of coil heating practice in steel mills. Flat Rolled
Products: Rolling and Treatment I, Metallurgical Society Conf., AIME, Interscience, New York,
T2 = 1649F. 1959.
_i__
Fig. 9.3 Steady-state temperature distribution in a hollow cylinder. T;
At this point, the reader should note that this problem and the one discussed in
Se~tion 7.4 are identical. By way of illustration, we have used a different starting
pomt here. In any event, the temperature profile is given by
T - T2 In (r/r 2 )
(7.65) To
T1 - T2 In (rifr 2)
2 Fig. 9.4 Steady-state temperature distribution in a composite cylindrical wall.
It then follows that the heat flux q, (Btu/hr-ft ) and the heat flow Q (Btu/hr) are
kT1 - T2
q = -- ' (7.66) Note that the form of the thermal resistance attributed to the cylindrical wall
r r In (rifr 2) differs from that of the slab. By comparing Eq. (9.17) with Eq. (9.12), we can
2nkL contrast the thermal resistances as follows:
Q = In (r / r ) (T1 - Tz). (7.67)
2 1
L
Infinite slab: Rt=-
kA
9.2.4 Composite cylindrical wall
Consider the cylindrical composite wall shown in Fig. 9.4. For steady-state In (r 2 /r 1 )
Infinite cylinder: R1= ' (9 .18)
conditions, Q is constant, and for a length L we write 2nLk
2nk 1 L
Q = h;(2nr 1 L)(I; - T1 ) =
ln(r 2 /ri)
(T1 -
.
T2 ) Surface to fluid: Rt = ~h
2nk 2L As an application of Eq. (9.17), consider the design of single-wall tube furnaces
= 1n (r / r ) (T2 - T3) = h 0 (2nr 3L)(T3 - T0 ).
to operate at I; internally while placed in some environment at T 0 . The radial heat
3 2
flow through the furnace wall is then given by Eq. (9.17) written for a tube composed
Again, the series concept is the most convenient method to relate the heat flow to
the overall temperature drop: of a single layer with conductivity k;
Q= 2nL(I; - T0 ) (9.19)
(9.17)
1 1 r2 1
- +-In-+--
hh k r1 h0 r2
-----~------------~ ----------------~---~-- ------- ----O=>""S'-...--'J-..,. .. - ..._,............. ,__.,... .. ~ ... ET.11........ _ _ . _ _ _._..."'"~-~~------
Ifwe examine Eq. (9.19), we see that as r 2 (the outer radius) increases, there is an
T situation of zero wall thickness. The existence of a critical radius shows that under
increasing resistance to radial heat conduction as embodied in the In term. Simul- some realistic conditions, and contrary to common expectations, the heat loss
taneously, however, as r2 increases, the outer cooling surface area increases as well. through a tube furnace can actually be decreased by decreasing the insulating wall
This dual effect suggests that there exists a particular outer radius for which the thickness.
heat loss (or gain) is a maximum. To examine this proposed effect, fix ri. and
determine that particular value of r 2 for which dQ/dr 2 = 0: 9.3 STEADY-STATE, TWO-DIMENS~ONAL HEAT FLOW
h
I
,.:_,-
""'"'
0"'
N
0.8 T=O 0.10 T 0 T=O
0.6
y =0-:"",- -.........- - ' ! '
I
x=O T=T0 x=L
Fig. 9.6 Temperature distribution in a semi-infinite plate.
0.4
For such a solid the temperature field is two-dimensional, and under steady-
state conditions must satisfy
-a 2 T a2 T
ax2 + ay2 = 0. (9.21)
0.2
I
The method of separation of variables consists of seeking product solutions in First, we examine the boundary conditions for x. For Eq. (9.31) to satisfy Eq.
the form (9.22), X must vanish at x = 0; therefore, C 1 = 0. Similarly, X must vanish at
T(x, y) = X(x) Y(y), (9.26) x = L to satisfy Eq. (9.23). Therefore
where X is a function of x alone and Y is a function of y alone. Introducing Eq. sin .AL= 0. (9.33)
(9.26) into Eq. (9.21), we have Equation (9.33) requires A.L = 0, n, 2n, 3n, etc., or, in general, An= nn/L,
2
d X 2
d Y where n = 0, 1, 2, 3, etc. The equation sin A.L = 0 is called the eigenfunction and
Y dx2 + X dy 2 = 0 (9.27) the values of An, the eigenvalues. We have now considered both boundary condi-
tions for x. At this point we may write
or, by separating the variables, . nnx
X = C2 sm- (9.34)
( (~)( ~:!l
-(~) ~:~) (9.28)
L
Equation (9.34) obviously satisfies (9.29a) for any eigenvalue; it is also true that the
sum of all the eigenfunctions satisfies Eq. (9.29a). Therefore, we write
The right-hand side ofEq. (9.28) is independent of x (by hypothesis); therefore, the
00
left-hand side is also independent of x, and hence must be equal to a constant. nnx
Similarly, the left-hand side is independent of y, and demands the right-hand side X =LC
n=O n
sm-
L
(9.35)
to be independent of y. Hence, it follows that both sides equal an arbitrary con-
When the boundary conditions for y are applied, we see that Eq. (9.24) requires
stant; this constant A. 2 is called the separation constant.* Therefore
C3 = 0 in Eq. (9.32). Thus,
d 2X
dx 2 + A. 2 X = 0, (9.29a) Y = C4e-ly = C4e-(nn/L)y. (9.36)
The product solution now takes the form
d2 Y
dy2 - A_2y = 0. (9.29b) 00
T = X. y = L
A e-(nn/L)y sin nnx' (9.37)
n=O n L
Equations (9.29a) and (9.29b) are homogeneous, linear equations with constant
where the An's have absorbed the constants involved.
coefficients. This type of equation can be solved by setting X = e"x and Y = ebX,
For the final boundary condition (Eq. 9.25), Eq. (9.37) becomes
and substituting in Eqs. (9.29a) and (9.29b), respectively. For Eq. (9.29a), a = iA.,
00
and the solution takes the form nnx
To = L
An sm - (9.38)
(9.30) n=O L
To determine all the various values of An, we multiply both sides of this equation
Making use of the identities e ilx = cos A.x i sin .Ax, the solution takes on a more by sin mnx/ L, where m is one particular integral value of n; we then integrate
commonly used form: between x = 0 and x = L:
mn(fHf) ~ f .~/
X = C 1 cos A.x + C2 sin .Ax. (9.31)
For Eq. (9.29b), b = A., and the solution is T, f
x/L=O
s;n
x/L=O
f)
s;n nn( s;n mn( f) a(f) (9.39)
l
L
~---~-~~-~~rnoo~~~~~~~-~--------~-~----------------~----,~~-,~~93---~~---~--------~-~reoo~~ffi~~~~~~~------
I
The final solution is Noting that A 0 = 0, we obtain the final solution
00
4 . mrx
L -e-(nrr/L)ySlll- (9.41)
(9.51)
n= 1 nn L
n,odd
[f(x) - T1 ]
.L
Sill
nnx
dx. (9.52)
B.C. 1 at x = 0, T= T1 ; (9.42) x=O
B.C. 3 at y = oo, T= T1 ; (9.44) In this section we go on to consider the finite plate with the three edges x = 0,
(9.45) x = L, and y = 0 maintained at zero temperature,
B.C. 4 at y = 0, T = f(x).
T(O, y) = T(L, y) = T(x, 0) = 0, (9.53a, b, c)
The development of Eq. (9.41) is possible due to the homogeneous boundary
conditions, namely, Eqs. (9.22) and (9.23), that is, the method depends on the and the fourth edge y = H maintained at a temperature distribution f(x),
temperature being zero at the two extremes of x. For the case at hand, we simply
T(x, H) = f(x). (9.54)
transform the temperature variable T into 8, where e = T - T1 . With 8, then, the
Laplace equation becomes Again Eq. (9.21) must be satisfied, and we employ the method of separation of
<J2e <J2e variables. Equations (9.53a) and (9.53b) will be satisfied if
(9.46)
ax2 + ()y2 = 0, X(O) = X(L) = 0, (9.55a, b)
and the boundary conditions are then whereas Eq. (9.53c) implies the condition
8(0, y) = 8(L, y) = 8(x, oo) = 0 and 8(x, 0) = f(x) - T1 = F(x). Y(O) = 0. (9.56)
Again, assuming product solutions, e = X Y, and proceeding in exactly the As before, when Eqs. (9.55a) and (9.55b) are satisfied,
same manner as before, we can obtain by referring to Eq. (9.37): 00
nnx
00
nnx X = L C sin- (9.35)
e= L A e-(nrr/L)y sin- (9.47) n=O n L
n=O n L
J:his time we choose to write the general solution of Eq. (9.29b) as
By applying the "toughest" boundary condition, we get
Y = C3 sinh A.y + C4 cosh A.y. (9.57)
00
nnx
F(x) = L An sin-, (9.48) (This is another form of Eq. (9.32) if the definitions of the hyperbolic sine and
n=O L hyperbolic cosine are substituted.) The boundary condition for Y, Eq. (9.56),
and proceeding as before, we have eliminates C4 , and as before, A.n = nn/ L. Thus it follows that the product solution
takes the form
x/L=O
(9.49)
T = X Y =
n=l
I A sinh nny sin nnx
n L L
(9.58)
Therefore, we evaluate the arbitrary constants: We must now determine the coefficients An in such a way that the remaining
x=L condition, Eq. (9.54), is satisfied. For this boundary condition, the solution becomes
An =
2
L f nnx
F(x) sin L dx. (9.50) 00
nnH) S
. innx
f(x) = L (
Anslllh- ll- (9.59)
x=O n=1 L L
-"--"-"-,~"_" __"-r....----"Y;<t-- 291
Following the same procedure as before to determine coefficients, we obtain
. nnH 2
An smh L - = L
J .
L
nnx
f(x) sm L dx. (9.60)
0
and writing Bn = An sinh (nnH/L), we find that the solution (Eq. 9.59) takes the
form x=O x=L
. h nny
00
sm -
L . nnx
T= I Bn sm-, (9.61)
(a) (b)
n=l . hnnH - L
sm - - Fig. 9.8 Transient temperature distributions during cooling of (a) a thick plate, and (b) a thin
.L plate .
where
-in Fig. 9.8(a). If the plate i~ thin (Fig. 9.8b), 'and/or its thermal conductivity high,
L
Bn = L2 Jf(x)smLdx.
. nnx (9.62)
then the temperature gradients within the plate are negligible, and we may consider
the temperature only as a function of time. In the next section, we shall examine
0 when this criterion is met; we find that if hL/k < 0.1, then the analysis that follows
is valid.
f(x)
For this situation, the heat balance takes a rather simple form. We equate the
'~}MGM]
rate of heat lost by the plate to the rate of heat transfer to the fluid:
dT
- VpCP dt = hA(T - TJ), (9.63)
y o i g(x) I
x=O x=L where V = volume of the plate, and A = area of the plate exposed to the fluid.
Fig. 9.7 Method of solving the temperature distribution in a rectangular plate with general Rearranging for integration, we have
boundary conditions, by adding the solutions of four simple cases.
T t
We can obtain the solution of the more general problem of dealing with the
JT- TI
dT
-vphAc
p
Jdt, (9.64)
temperature prescribed along all four edges by superimposing four solutions, such T; 0
as the one obtained here, each corresponding to a problem in which zero tempera-
tures are prescribed along three of the four edges. Figure 9.7 schematically depicts T- TI= exp(-hAt) (9.65)
this method of solution. Ji - TJ pCPV
The analysis applies only to the cases in which internal gradients are negligible.
9.4 TRANSIENT SYSTEMS, FINITE DIMENSIONS
If this requirement is met, we speak of Newtonian cooling. Note that no geometric
In this section, we consider conduction heat transfer in solids, in which the tem- restrictions are imposed, since Eq. (9.65) just specifies A and V; also note that the
perature varies not only with position in space, but also undergoes a continuous solution does not contain the conductivity of the metal.
change with time at any position. In particular, this topic is of vital importance in The temperature history of such objects during cooling is illustrated in Fig. 9.9,
quenching or cooling operations. which shows that for the same surface radius, R, or semithickness, L, the time it
takes an infinite cylinder (or infinite square rod) to cool within 1 % of thermal
9.4.1 Newtonian heating or cooling equilibrium with its surrounding fluid is almost 50 % longer than for a sphere
When a solid such as a flat plate, initially at a uniform temperature T;, is cooled by (or cube) of the same material. On the other hand, an infinite plate of the same
a fluid at temperature TJ, the temperature distribution varies with time, as shown material and semi thickness equal to the cylinder radius requires 100 %more time.
---r,n111'-"1CUr~31.'Clll~lllH.C-UHllC'IUHUllS >J
For a plate or slab whose thickness is much less than its width and length, A/V may
be assumed equal to A/V for an infinite plate. Therefore
A 12 in. _1
V L i .
gill.
Ift = 96 ft
dT (90)(96) .
0.1
Infinite plates a) dt = (l 0)(0. 24) (600 - 200) = -80,000 F/hr
8
0
or
dT
-22.2F/sec at T = 600F.
dt
45 96
b) dT = ( 0)( ) (300 - 200) = -100 000F/hr
0.01
dt (180)(0.24) '
or
dT
-27.8F/sec at T = 300F.
dt
c) First calculate time ti to reach 500F. Use Eq. (9.65):
0.001 0
2 3 4 _T_-_T-"-J = exp[--h-~ti]'
ht/CppR or ht/CPpL 7; - T1 pCP V
Fig. 9.9 Cooling (or heating) of bodies with negligible temperature gradients. R is radius; and
L is semithickness.
500 - 200 [ (90)(96)
1000 - 200 = exp - (180)(0.24) ti
J
Example 9.2 A plate of aluminum with dimensions tin. x 12 in. is solutionized Solving for t1' we have
at 1000F, and then quenched in water at 200F. Over the temperature range
1000-500F, the heat-transfer coefficient for quenching in the water may be ti = 0.00491 hr.
assumed constant as 90 Btu/hr-ft 2 F. In the range 500-200F, the heat-transfer Calculate time t 2 to cool from 500 to 250F. Now 7; is interpreted as 500F:
coefficient may be approximated as 450 Btu/hr-ft 2 F.
a) Calculate the cooling rate during quenching at that instant when the plate is
250 - 200 [ (450)(96) J
500 - 200 = exp (180)(0.24) tz
at 600F.
Solving for t 2 , we get
b) Repeat, when the plate is at 300F.
tz = 0.00184 hr.
c) Determine the time it takes the plate to cool to 250F. The properties of the
aluminum are: Therefore total time to cool to 250F is given by
k = 45 Btu/hr-ft F,
Cv = 0.24 Btu/lbm,F,
t = ti + t 2 = (0.00491 + 0.00184) hr
or
p = 180 lbm/ft 3 .
Solution. We can write an expression for the cooling rate by simply rearranging t = 24.2 sec.
Eq. (9.63): We should point out that the critical assumption embodied in Newtonian
dT h A cooling is that the internal temperature gradients are non-existent. The reader is
- = - - ( T - T1 ).
dt pCP V assured that the assumption, as applied to this particular example, is valid.
l
----- --=-=~~~~~~m~'"mre ~dionm - --Todifiration ;, ~'ilyrnoo~ired; 'P~ilira~,~~m~~~=~h=~~=-
For the flat plate of Fig. 9.8(a), the applicable differential equation by reference to J tial equation and boundary conditions become
Eq. (9.2) is I oe o2 e
oT o2 T I at= a ox 2, (9.71)
at = a ox2. (9.66) 111
:1 for mathematical purposes was homogeneous, that is, T = 0. Similarly, for and for Eq. (9.77), the solution is
mathematical convenience, Eqs. (9.68) and (9.69) should be homogeneous. In
G = exp(-A 2 a.t).
' summary, we state the three basic types of boundary conditions that can be treated
analytically: Boundary condition (9.73) requires that c 2 = 0, and Eq. (9.74) can be shown to
require that
(9.79)
T(boundary) = 0, (9.70a)
(9.80)
oT
ox (boundary) = 0, (9.70b)
Equation (9.80) is analogous to Eq. (9.33) in that An takes on an infinite number of
eigenvalues. Figure 9.10 indicates this for the first three eigenvalues.
or
Taking the product of Eqs. (9.78) and (9.79) with c 2 = 0, and realizing that all
oT h values of An satisfying Eq. (9.80) are suitable, we have
ox (boundary) k T (boundary) = 0. (9.70c)
00
Bi k/hL J. 1 L J. 2 L J. 3 L J. 4 L
cot XL
or 100 O.Ql 1.5552 4.6658 7.7764 10.8871
k
hL (XnL) 10 0.10 1.4289 4.3058 7.2281 10.2003
1 1.0 0.8603 3.4256 6.4373 9.5293
0.1 10.0 0.3111 3.1731 6.2991 9.4354
still remains to be satisfied. When Eq. (9.72) is substituted into Eq. (9.81), we get
. ctt
ei =
00
l
'I L,
2
exp (-Anett) COS (A.nx), (9.85) a) Find the time it takes for the center to reach 100F.
fJ; T; - TI n= 1 A.nL + sin (A..L) cos (A.nL)
b) When the center reaches 100F, what is the surface temperature?
I
J
' where the A.n's are the roots of Eq. (9.80), given in Table 9.1.
i Graphical evaluations of Eq. (9.85) and the analogous solutions for infinitely c) What would be the minimum possible time for the center of the bar to reach
long cylinders and spheres have been presented in many graphical forms for 100F if it were cooled in an ideal quenchant (H = oo) at 80F?
practical use. All these graphical presentations are given in terms of a relative 2 H.P. Gurney and J. Lurie, Ind. Eng. Chem. 15, 1170-1172 (1923).
temperature as a function of the Fourier number, Riot number, and relative position. 3
M. P. Heisler, Trans ASME 69, 227-236 (1947).
~ 1.0
I
s
2 0.8
b
0.6
0.4
0.2
Fo = Oll/L 2
(d) x/L = 1.0 (surface)
0 0.01
50 100 Fig. 9.11 (continued)
(a) x/L = 0 Fo =a t/L 2
1.0
~ 1.0
r,...~
I
I h"
~
~ 0.8 h' 0.8
r,... I
r,...
b 0.6 0.6
0.4 0.4
0.2 O.~
0
0 O.Ql 0.01 0.05 0.1 50 100
50 100
2 Fo=at/R 2
Fo =a t/L (a) r/R = 0 (center)
(b) x/L = 0.3
~ 1.0
I Cylinder
~ 0.8
~
b 0.6
0.4
0.2
.
l
~
,l
0
0.001
Fig. 9.11_ Temperature resl?onse ofan.infinite plate initially at a uniform temperature 7;, and Fig. 9.12 Temperature response of an infinite cylinder initially at a uniform temperature 7;,
then subjected to a convective environment at Tr. and then subjected to a convective environment at Tr.
;:s:. 1.0 ;:s:.
I I
,.:;-
~ 0.8 ~ 0.8
~ ;:s:.
,.!,
~ 0.6 b
0.4 0.4
0.2 0.2
;:s:. 1.0
The thermal properties of the stainless steel are k = 9 Btu/hr-ft F and rx = 0.158
I ft 2 /hr.
,.:;-
~ 0.8 Solution
,.!,~ 0.6 a) When T = I00F at the center, we have
T- TI 100 - 80
T; - TI = 400 - 80 = 0.0625
0.4
0.2
Also
0 Bi = hR = (25)(5/24) O
0.01 0.05 0.1 0.5 I 5 10 50 100 500 1000 k 9 = .579 .
2
Fo=oo/R
(b) r/R = 0.4
From Fig. 9.l2(a), rxt/R 2 ~ 3.l. Then
(3.1)(5/24) 2
t = (0.1 ) = 0.84 hr
;:s:. 1.0 ..,__---' 58
I or
~ 0.8
;:s:. t = 50.5 min.
b 0.6 b) For the surface temperature, we refer to Fig. 9.12(b). When Fo = 3.1 and
Bi = 0.579, we find that
0.4
0.2
i '
Therefore
50 I 00 500 1000
Thus Try L = 0.1 ft. In this case, Bi = 0.25 and, from Fig. 9.1 l(a), Fo should be 6.6.
at/R ~ 0.57,
2 However, with L = 0.1 ft, Fo = 5.36, so that this value of L does not satisfy Fo and
Bi simultaneously.
and Try L = 0.05 ft. Now Bi = 0.125, and Fo should be 12. With L = 0.05 ft, Fo
(0.57)(5/24) 2 is calculated as 20.
t = = 0.179 hr, We have bracketed the actual value, and after more trial and error, the value
0.158
or of L = 0.092 ft leads to agreement between Bi = 0.23 and Fo = 6.35.
As a check, the times to reach various temperatures with L = 0.092 ft should
t = 10.7 min. be calculated to see that the corresponding time-temperature curve does not
Example 9.4 The process of austempering requires that a steel be quenched to just intersect the nose of the TTT curve. The result of this calculation is shown in the
, ~, above its Ms temperature, and then isothermally transformed to a lower bainite figure.
structure that resembles tempered martensite. For a steel with the TTT diagram Thus we can say that a plate 2.2 in. thick can be fully austempered. Strictly
below, estimate the maximum thickness of plate that can be completely a us tempered speaking, since the TTT diagram is developed using small samples isothermally
by quenching into molten salt at 400F from a l 600F austenitizing temperature transformed, the "nose" of the curve should be moved slightly to the right for
with h = 50 Btu/hr-ft 2 F. For the steel, assume a = 0.46 ft 2 /hr and k = continuous cooling applications, but as a first approximation, the isothermal
20 Btu/hr-ft F. TTT diagram may be used.
;;-. 1600
f..; ~------
A1
-----(0 In conclusion, it is important to indicate when limiting cases are valid, such as
1000
' ',
A ' ,
......
Newtonian cooling. This, of course, simplifies our work, as Fig. 9.14 demonstrates.
Figure 9.14 presents the solution for the infinite plate such that temperature
distributions within plates are shown for different times during cooling in media of
}
500
' __~
various Biot numbers. Examination of Fig. 9.14 shows that the temperature
gradients within the slab decrease as Bi decreases. A low value of Bi, physically
interpreted, reflects low resistance to heat flow within the body, L/k, relative to the
w.c1 200"-----------'l.--M, resistance of the cooling media h- 1 . In practice, when Bi < 0.1, the temperature is
:~ ~:--
~ <>' . 10 102 10 3 104 nearly uniform within the plate. For such cases, we approximate the cooling or
1
'<.; "' t .sec ) heating processes as being controlled solely by surface resistance, and we can apply
Since we must cool at such a rate ~s to bring the center. of the plate ptst the the Newtonian relationship, Eq. (9.65), as a very close approximation. We may also
"nose" of the curve without undergomg any transformation, we can start by apply the same approximation to the heating or cooling of cylinders and spheres,
assuming that point P is the critical point. This can be tested later. Since point with the criterion Bi < O.l still being in effect.
P is at 700F, then At the other extreme, when the 'Cooling or heating process can be considered to
T - TI = 700 - 400 = 0. 25 . be completely controlled by the internal resistance of the body, then the surface
T; - TI 1600 - 400 temperature, in effect, immediately changes to TI, the temperature of the fluid, and
remains constant at this temperature. This situation can be considered as a special
Solution. Figure 9.ll(a) gives us the required result. What is needed, is the value
case with Bi~ ro. Alternatively; a new solution to Eq. (9.66) could be developed
of the semithickness L that will result in agreement between Bi, Fo, and reduced for the same boundary conditions as those ofEqs. (9.67) and (9.68), with T(L, t) = IJ-
temperature. Since replacing Eq. (9.69). The solution to this problem is given by
t ~ 420 sec, or 0.116 hr,
T- TI_=~ ~ (-1)". exp [-(2n + 1) n !!:!._]cos (2n + l)n ~.
2 2
then 9 86
2 2
Ii - TI n /:o 2n + 1 4 L2 2 L ( )
Fo = (0.46)(0.l 16)/L = 0.0536/L
i.
Examination of Fig. 9.14 shows that the criterion, T(L, t) = TI, is closely
and
approximated when Bi= 100. Thus when Bi > 100, Eq. (9.85) can be very closely
t 'i
B1. = -
50L = 2.5L. approximated by Eq. (9.86).
20
9.4.3 Long-time and short-time solutions
The solutions to many problems of unsteady-state heat conduction are in the form 0.80
of infinite series, such as Eqs. (9.85) and (9.86). This type of series converges rapidly
for large Fourier numbers. For short times (small rxt/ L 2 ), however, the convergence
is very slow, requiring an expansion of a great number of terms in the series to
obtain sufficiently accurate answers. If we require a solution for short times, OAO
alternative series that.evolve by the method of Laplace transforms when solving
Eq. (9.66) are more convenient. These alternative series have the advantage of
converging rapidly for small rxt/L 2 , but the disadvantage of converging very slowly
for large rxt/L 2 . Thus the two kinds of solutions complement each other, depending 0 ~-~~~--~-~~~~~----"__JO.SO
on what value of rxt/L 2 is of interest. These considerations are important when om o.o5 0.1 o.5 1 2 4 6 10 50 100
Bi
we wish to use the actual equations, rather than the graphical solutions, as might
be the case in computer programming, for example. Fig. 9.15 Values of constants for Eq. (9.87).
l
If we desire a long-time solution, rather than the complete series of Eq. (9.85), that Eq. (9.87) is converging rapidly, and therefore a good answer results by use of
it is convenient to use only the first two terms of the series. For this case, we write this equation. Solving, we have
T-T
T
,
_ jJ = 2 L n= 1
2
Examp~,5 \.\ slab of aluminum 4 in. thick at 975F is quenched in a bath of A difference of only 8F lies between the two methods. This difference would have
water at 75F. The heat-transfer coefficient is estimated to be 2000 Btu/hr-ft 2 F. been smaller if we had considered longer times, but would have been larger for
shorter times. Usually, Eq. (9.87) suffices for problems of this type if Fo > 0.25;
a) Calculate the temperature at the center of the slab 30 sec after being plunged it is wise, however, to compare results with those obtained by using Figs. 9.11-9.13.
into the water. Use Eq. (9.87).
When the long-time solution is not appropriate, we may utilize the alternative
b) Repeat, using Fig. 9.11.
series mentioned above. Here we only present the first several terms of the series
The properties of the aluminum are the same as those listed in Example 9.2. without going through the analysis of solving the differential equation. We may
deduce the following expression from an equation given by Carslaw and Jaeger 4
Solution
~2 = ~0.187. Equation (9.88) contains the complementary error function, erfc. The erfc N
is related to the erf N (the error function) simply by
Also erfc N = 1 - erf N,
Fo = ~ _ (45)(30/3600) _
and the erf N is defined as the value of a definite integral. The definite integral is
L2 - (180)(0.24)(2/12) 2 - 0. 3 13.
N
4
H. S. Carslaw and J.C. Jaeger, Conduction of Heat in Solids, second edition, Oxford University
The first term within the brackets is much larger than the second; this indicates Press, 1959, page 310.
__________ _______________ ______
,, ,.
-----
--------~-~~r---
------~-~-~------m""'""""""~
The error function is commonly encountered, and hence it has been tabulated. An exp [Bi(l - x/L) + Bi 2 Fo] = exp [7.41 + 7.41 2 x 0.125]
abbreviated compilation is given in Table 9.2.
= 1.55 x 106 .
Table 9.2 Tabulation of the error function
N erf N N
erfc [Bifto +
1
-;] =
2v Fo
erfc [7.41 Jo.i2s + 2vh]
0.125
erf N N erf N
= erfc [4.04] "'i 0.230 x 10- 7 _
0.00 0.00000 0.50 0.5205 1.0 0.8427
0.05 0.05637 0.55 0.5633 1.1 0.8802
Therefore
0.10 0.1125 0.60 0.6039 1.2 0.9103 T- T1
0.15 0.1680 0.65 0.6420 1.3 0.9340 --- = 1 - (2 x 0.0452) + (2 x 1.55 x 106 x 0.230 x 10- 7 ) = 0.981,
~)
0.20 0.2227 0.70 0.6778 1.4 0.9523
Ii - T1
0.25 0.2763 0.75 0.7112 1.5 0.9661 T = 0.981(975 - 75) + 75 = 924F.
0.30 0.3286 0.80 0.7421 1.6 0.9763
0.35 0.3794 0.85 0.7707 1.7 0.9838 As the reader probably suspects, long-time and short-time solutions are also
0.40 0.4284 0.90 0.7969 1.8 0.9891 available for cylinders and spheres. These solutions can be deduced by reference to
0.45 0.4755 0.95 0.8209 1.9 0.9928 Carslaw and Jaeger. 5
1.00 0.8427 2.0 0.9953
9.4.4 Cooling and heating rates
Notes
N In metallurgy, the rates of heating and/or cooling are often more important than
a) erf N =.]; J e-P
2
d/3, the determination of the temperature itself. For cases where Newtonian cooling
applies, the determination of cooling rates is not too difficult; we discussed this in
0
Section 9.4.1, and illustrated it in Example 9.2. However, when Newtonian cooling
b) erfO = O; erf oo = 1, does not apply, it is necessary to resort to other means. If analytical expressions are
c) erfc N (complementary error function) = 1 - erf N, desired, then solutions such as those discussed in Section 9.4.3 may be used to
determine oT/ot. If approximate and graphical solutions are needed, then the
2N
d) N < 0.2, erf N ~ _, charts in this section may be consulted. Figures 9.16, 9.17, and 9.18 include these
Jn charts for infinite plates, infinitely long cylinders, and spheres, respectively. In
e-N2 order to save space, we do not present the surface rates for cylinders and spheres.
e) N > 2.0, erfcN ~ ~, They are almost the same as those for infinite plates; for all values of Bi, the surface
nN
co'oling rates for spheres are only 3-4 % higher than for plates. However, the
f) erf(-N) = -erf(N). cooling or heating rates of plates, cylinders, and spheres do differ greatly in the
internal positions.
Example 9.6 Repeat Example 9.5, but calculate the temperature at the center of Example 9.7 Determine the cooling rate at the center of the cylindrical bar,
the slab 12 sec after being plunged into the water. described in Example 9.3, when the center reaches 300F.
Solution Solution. From Example 9.3, the Biot number is 0.579. For the center, we have
T- T1 300 - 80
Fo = (0.313)(~~) = 0.125. I; - T = 400 - 80 = 0. 688
1
L
-~------r \.
00
'
0 0
0 0
N
0 "
0 0"'
00
0
fl~e
1.0 0 2.0 0
I I o e
h &:.-
-.=o--1.-0 0.9
"'
1.8 h-------
Use top and h'"'
-li:Q '"'/ ~ 3.6.....---~---::;.....~~~--~--~-~-.--,
left scales I I
0.8 1.6~ "5'
"" h'"' Ih'"' 3.2
--------
0.7 1.4 I I
h h- 2.8
0.6 1.2
0.3
0.8
0.2 0.4
0.4
0.1 0.2
0 N 00 ~ "1':
0
0
"0
0 "'
0
0
0
0
0 0 0
Fo
(a) x/L = 0 (center)
Fig. 9.16 Rate of temperature increase of an infinite plate initially at a uniform temperature,
and then exposed to a uniform-temperature convective environment. (From V. Paschkis,
Welding Research Supplement, Sept., 1946, pages 497-502.) 0.4
0.5
0.6
1
0.7 ffi=lO
Then by using Fig. 9.17(a), we arrive at
0.8
a( T- T1 )
0.9
T;-TJ ~
1.17.
Bi o(Fo) J.0 L---N~--._,....,____.'_oo~-----'N:--.---':._,.-~'0:---00"':-:'_
0 0 0 q q 0 0 ci ci 0
Therefore o o o o o Fo
0 28 I
f::, m=ffi=oo
---~1~
""
24
~~-
- ""
-li:i5
20
5.6 I .
Bl= 0
16 4.8
4.0
12
1.6
0.8
0
Fo
(c) x/L = I (surface)
0.2
""" \0 00
0.6
0 0 o- N """ \0 00
0 4.0 ci ci ci ci ci 0 cici-
2.0
i.... 0 ~1-0.8
'15' i....
-=::::::::.. 3.6 1.8 -'.2:_ I 11 ~
h...,lh..., h "-f-:<-1 it:" 1.0
I I 3.2 1.6 h...,lh...,
hf-:," I I
-li:i5 1.2
~2.8 1.4 h f-:,"
""
-li:i5 I .4
2.4 1.2
1.6 0.8
1.8
1.2
Use bottom and
0.6 rh= 10
2.0
right scales rl ~ '-.q 00
0.8 0.4 ~ 0 """
0 '
~
00 -
~6
rl
a 6 0 ci
a 0 ci 0 0
Fo
0.2 (b). r/R = 0.5
Fo
(a) r/R = 0 (center)
Fig. 9.17 Rate of temperature increase of infinitely long cylinders initially at a uniform
temperature, and then exposed to a convective environment at constant temperature. (From
V. Paschkis, Welding Research Supplement, Sept., 1946, pages 497-502.) For the surface refer
to Fig. 9.l6(c).
N
0
... '
00
oqo N 'tj"\OOO
6.0 ,=-~~or~~o~or_or~~-or~~~o-~or-o~--~~~~~~~~~~ 3 .0 ~
6
0
~
......__
'15""
5.6 ~
2.8 h"I h"
" h"lh"
hI I 5.2
h-
2.6~
I I
~ 4.8
Use top and 2.4 - "'IPS
left scales
4.4 2.2
4.0 2.0
3
3.6 1.8
3.2 1.6
2.8
2.4 1.2 0
~
~ 0.4
2.0 1.0
1.6
Use bottom and
right seal es 0.8
h"I h-h" 0.8
h
I I
0.8 1.6
0 '---~-'-~--'----'---'-~~-'-~~J,,----,1:=;!""""~~--~~-'----'~'---'o 2.4
"o::t N o::f: \Cl 00.,....,
~ 0 0 c:i 0
Fig. 9.18 Rate of temperature increase of spheres initially at a uniform temperature, and then (b) r/R = 0.5
Fo
exposed to a convective environment at constant temperature. For the surface refer to Fig.
9.16(c). Fig. 9.18 (continued)
- - - - - - - - - - ---- --------------~--~-r..---,,.,", __
I
9.5 TRANSIENT CONDITIONS, INFINITE AND SEMI-INFINITE SOLIDS various T;n's can be considered to form a function of x', f(x'). In this case, examine
In Section 9.4, all the solids we considered had at least one dimension of finite the series in the limits as the slices all approach zero thickness
00 2
extent. In this section, we consider the so-called infinite and semi-infinite solids. f(x') [-(x - x')
The solutions we shall deal with can be applied when the time involved in transient T(x, t) = Jim L r.::::; exp
]
Ax'. (9.91)
~x~on;l 2v'nr:xt 4r:xt
situations is very short, or during the time of interest the depth of material affected
by the boundary condition is less than the thickness of material itself. There will be As Ax' -+ 0, the infinite summation becomes an integral
0
T
T
T; T;
- -6.x'
Fig. 9.19 Temperature distribution in an infinite solid showing (a) an initial temperature peak,
As an example consider an initial distribution (Fig. 9.20). For this case, we
and (b) the decay of the peak with time. develop the solution by recognizing that f(x') = 0 for all x except a < x < b
where f(x') = T;. Substituting this information into Eq. (9.92) yields
I +oo I
f/o
9.5.1 Infinite solid a b
Consider the infinite solid in Fig. 9.19(a) where at time zero a thin slice of material
T = ] dx' + dx. (9.93)
Ax' thick at x = x' is at some temperature T;. This temperature peak decays with -oo
I
a b
1,
time, as shown in Fig. 9.19(b). Again for transient heat conduction, Eq. (9.66)
applies, and the solution is
Let f3 =(x' - x)/2fa; this transforms Eq. (9.93) into
(b-x)/2..;af ( (b-x)/2,/ai (a-x)/2./iil )
T(x, t) = b
2v' mxt
exp [
2
(x - x') ]Ax'.
4ca
(9.89) T = T;
Jn
J
fJ;(a-x)/2./iil
e - pi d/3 = T;
2
~
Jn
J
0
2
e - P d/3 - ] ; J
0
2
e - P d/3 .
If another slice of material of the same thickness exists at time zero with a (9.94)
different temperature T;, then the temperature at x and tis the sum of contributions Therefore we write Eq. (9.94) as
from both peaks. For this case then, Eq. (9.89) applied to the peaks at x'1 and x~ is
given by
2 2
T = ; T[ erf
2
- x) - erf (a fa
(b fa
2
- x)] (9.95)
~J f(x'){exp [-(x -
00
2 2
8 = x') ] - exp [-(x + x') ] }dx'. (9.106)
2v nt:J.t 4t:J.t 4t:J.t
0
Next we change variables, f3 = (x' - x)/2fa and {3' = (x' + x)/2fa, and
substitute f(x') = 8;:
(b) 00 00
8
Fig. 9.21 Temperature distributions in infinite solids that satisfy boundary conditions for (9.107)
semi-infinite solids. (a) Odd function, T(O, t) = 0. (b) Even function, 8T(O, t)/8x = O.
We develop the solution to this problem by referring to an infinite solid with the or noting that primes are no longer necessary, we have
initial condition depicted in Fig. 9.2l(a). For the odd function where f(-x') = 00
f -~)exp f
00
2
T = [-(x - x')2] dx' + f~ exp [-(x - x') ] dx'. (9.98)
- 00
2 nt:J.t 4t:J.t
0
2 v nt:J.t 4t:J.t
-x/2
f ../iI
The problem with the flux equal to zero at the surface is set up in a similar
manner. For this case, the boundary conditions for the region 0 :::; x < oo become
T(x, 0) = f(x); (9.99) -x
2V<if ~
iJT ~ +x/2-.1~
ox (0, t) = 0. (9.100)
2..;at f
We indicate the method of setting up this situation in Fig. 9.2l(b), where an
Fig. 9.22 Schematic representation of the integral in Eq. (9.108).
even function, f(x') = f( - x'), gives symmetry to the temperature field about
x = 0, so that condition (9.100) is automatically satisfied. Thus we use Eq. (9.92) Figure 9.22 schematically ind!cates these integrals, and shows that their sum
in the following form results in
f f
0 + 00
T = f( -.x') exp [- (x - x')2] dx' + f(x) exp [-(x - x')2] dx'. (9.101) 8
2~ . 4t:J.t 2~ 4t:J.t (9.109)
-oo 0
Now as an application of Eq. (9.98), consider the important and often en-
countered problem of the semi-infinite solid with the boundary conditions The solution in its final form is
T(x, 0) = I;; (9.102) T- T x
= erf--
_ _ _s (9.110)
T(O, t) = T,. (9.103) T-T
' s r;(
2 y<l.l
~---------------------------- ------- - - - - - - - -.. . --r-----..----------r------------------
where
y=~fa[
k
~+~fa].
vat k
(9.114)
In this section we present methods of obtaining solutions for solid shapes such as
cubes, rectangular bars, and short cylinders. These solutions can be obtained
directly by simply combining solutions for the semi-infinite solid, the infinitely long
cylinder, and the infinite slab, which are given in Sections 9.4 and 9.5.
As an example, consider an infinitely long bar with a rectangular cross section
2L by 21, using coordinates as illustrated by Fig. 9.24. The bar is initially at the
x
uniform temperature I;, and then it is suddenly exposed to a convective environ-
2-./W ment at T1 . In such a bar, T(x, y, t) must satisfy the partial differential equation
- '). .l .;lJ
Fig. 9.23 Temperature history in a semi-infinite solid with surface resistance. (From P. J.
J ar = (?PT ?J2r)-
I (9.115)
~t +!~2
Schneider, Conduction Heat Transfer, Addison-Wesley, 1955, page 266.) 2
:. (X ax_ J
T(x, 0) = I;; Example 9.8 A short cylindrical bar of stainless steel, 5 in. in diameter and 6 in.
(9.111)
long, is heated to 400F. The bar is then cooled in a blast of fan-forced air at
k oT~~, t) = h[T(O, t) - T1 ]. (9.112)
80F with h = 25 Btu/hr-ft 2 F. Use the same thermal properties as in Example
9.3.
We present only the final solution: a) After the bar has been cooled for 10 min, what is the temperature at its.geo-
metrical center?
T-T;
-
T1 - I; 2fa
(x}
- - = erfc - - - eYerfc - - + -y1rxt ,
2fa k
[x he] (9.113) b) What is the surface temperature midway between the ends after 10 min of
cooling?
------
rI
----~-~-- - - ---~,~unp1e-mun.1u11na1~1u11in-pruu1~RIS---~z-
xi
1 I
T 21
Infinite reel. bar
C(r,t)
Semi-infinite reel. bar
C(r,t)S(z,t)
Reel. parallelepiped
C(r,t)Z(z,t)
J..
z
J
'i
.,.... - -+----.._
Fig. 9.24 Rectangular bar showing the system of coordinates. Fig. 9.25 The composition of product solutions.
l
- ------------------,-------
Solution PROBLEMS
a) From Example 9-3, 9.1 A furnace wall is constructed of 7 in. of fire brick (k = 0.60 Btu/hr-ft F), 4 in. of red brick
(k = 0.40), 1 in. of glass-wool insulation (k = 0.04), and tin. steel plate (k = 26) on the outside.
. hR The heat-transfer coefficients on the inside and outside surfaces are 9 and 3 Btu/hr-ft 2 F,
BIR = k = 0.579, respectively. The gas temperature inside the furnace is 2500F and the outside air temperature
is 90F.
at
FoR = Rz = 3.1, a) Calculate the heat-transfer rate through the wall (Btu/hr-ft 2 ).
b) Determine the temperatures at all interfaces.
and
9.2 Consider the flow of heat through a spherical shell. For'steady state conditions, the
C(O, t) = ( To - TI) inside surface (r = R 1 ) is at temperature TI> and the outside surface (r = R 2 ) is at T2 .
= 0.0625.
T; - Tf inf.cyl. a) Write the pertinent differential energy equation that applies.
Additionally, b) Write the boundary conditions and develop an expression for the temperature distribution
in the shell.
. hL (25)(3/12) c) Develop an expression for the heat flow (Q, Btu/hr) through the shell.
BIL =h = 9 = 0.695,
9.3 A semi-infinite plate with edges at x = 0, x = L, y = 0 and y = oo is subjected to the
Fo _ at _ (0.158)(0.84) _ following boundary conditions at steady state.
L - L2 - (3/12)2 - 2.13,
x = 0, T= 0
and from Fig. 9.1 l(a),
x = L, T= 0
X(O, t) = (
T,0
T; -
-T)TJ I = 0.31. y = oo, T= 0
inf.plate nx
y = 0, T = TA sin L (TA = constant).
The answer we seek is found by using the product solution.
T- TI a) Write an expression for T(x, y).
T; _ TI= C(O, t) X(O, t) = (0.0625)(0.31) b) Write an expression for the heat flux along the edge y = 0.
9.4 The temperature Tis maintained at 0F along the three sides of a Jong bar with a square
= 0.0194. cross section (I ft x I ft). The fourth side is maintained at 100F.
Then
a) At steady-state, find an expression for the temperature Tat any point (x, y) in the bar.
T = (0.0194)(400 - 80) + 80 b) Calculate the rate at which heat is transferr.ed through the hot face (k = 50 Btu/hr-ft F).
= 86.2F. Expr,ess your answer in Btu/hr.
c) What is the temperature at the center of the bar?
b) For this case we use Fig. 9.12(b):
9.5 A thin wire is extruded at a fixed velocity through dies, and the wire temperature at
the die is a fixed value T0 The wire then passes through the air for some distance before it is
C(R, t) = (TR - TJ) = 0.040.
rolled onto large spools where the temperature has been reduced to TL. It is desired to investi-
T; - TI inr.cy1.
gate the relationship between wire velocity and the distance between the extrusion nozzle and
T- T roll for the specific values of T0 and TL.
T; _ ~ = C(R, t) X(O, t) = (0.040)(0.31)
= 0.0124. x=L
Then
T = (0.0124)(400 - 80) + 80
= 84.0F. x=O
------------:a_-.-.._..,.,."',..I'11------;:ll'"M'"1------
- - - = exp [ - ( /3
T-T,,
To - T,,
- N)h)J
pi + -
Rk
x ;
v
/3 =-
21)(
9.9 An open-ended cylindrical section of a pressure vessel 10 ft in diameter with 8-in thick
walls is being heat-treated The wall temperature is brought to a uniform value of 1750F.
Then the vessel is quenched in slow oil at 70F.
9.6 Steel b,!111 be.!lrings (}in. diameter) are austenitized at 1600F and then quenched in a
large tank of oil at 100F. Calculate: a) How long does it take for the surface to reach I000F?
b) What is the temperature at the center of the wall at that time?
'' . a) The time to cool the center of the bearings to 400F.
b) The surface temperature when the center is at 400F. 9.10 A cylindrical piece of steel 2 in. in diameter and initially at l 600F is quenched into 70F
q:) The space-mean, temperature when the center is at 400F. water (H = 18.0). Calculate the temperature 0.2 in. from the surface of the piece after 1 min,
d) If 10,000 balls are quenched per hour, calculate the rate of heat removal from the oil that 2 min, and 5 min. Compare your results with the temperature at the same location if the
must be accomplished in order to maintain its temperature at 100F. piece had been quenched in oil (H = 6.0).
Data: 9.11 Compute the temperature, as a function of time, across a slab of steel 4 in. thick, cooled
h = 300 Btu/hr-ft2 F, from 1600F by water sprays from both sides. Repeat for aluminum.
p = 450 Jbm/ft 3 , 9.12 Consider a short cylinder 6 in. high and with a diameter of 6 in. The cylinder is initially
at a uniform temperature of 500F and cools in ambient air at 80F.
cp = 0.15 Btu/lbm F,
a) Write the partial differential equation that describes the temperature within the cylinder.
k = 25 Btu/hr-ft F. b) Calculate the temperature at the geometric center after 1 hr of cooling.
c) Calculate the temperature on the cylindrical surface midway between the end faces after
9.7 Copper shot is made by dropping molten droplets into water at I00F. The droplets may
1 hr of cooling.
be approximated as spheres with a diameter of 0.2 in. Calculate the total time for the droplets
to cool to 200F if they enter the water at 2200F. d) In answering parts (b) and (c), show why your calculation procedure was justified, that is,
demonstrate that the differential equation in part {a) is satisfied.
Data for Cu, in units of Btu, lbm, ft, F, etc.:
9.13 A steel blank, 1 ft in diameter and 2 ft long, is heated in a preheating furnace maintained
Freezing point = 1985F
at 2080F.
cp (solid) = 0.09 a) Calculate the temperature in the center of the blank after the blank has been heated for
2 hr from an initial temperature of 80F.
cp (liquid)= 0.12
b) Calculate the time required to heat a smaller blank, t ft in diameter and 1 ft long, to the
Heat of fusion = 89 same center temperature as the larger blank in Part (a).
p (solid) = 560 Data:
p (liquid) = 530 .h = 20 Btu/hr-ft2 F,
Temperature range 9.14 The temperature field T(x, y, t) in an infinitely long rectangular (2L x 21) bar must
h, Btu/hr-fti F
satisfy the partial-differential equation
2200-1200F 80 air air 1 ar
1200-200F 400
I -
axi+ -ayi= -I)( -at
j
Prove that T(x, y, t) can be found by the product
T(x, y, t) = T1 (x, t) Tdy, t),
where T1(x, t) is the solution for the temperature history in the semi-infinite plate bounded by
- l < x < +/,and TL (y, t) is the solution for the temperature history in the semi-infinite plate 10
bounded by -L < y < L.
9.15 In the flame hardening of surfaces of thick (semi-infinite) steel parts a very hot flame is
played on the surface for a short time and then a water quench follows directly. If the surface
SOLIDIFICATION HEAT TRANSFER
of a steel part is brought essentially instantaneously to 2400F, and the flame I in. wide moves
at a rate of 1 in. per min with a water quench following directly, to what depth can a steel with
~,the continuous cooling transformation diagram given below be hardened to a 100% martensite
structure?
c-:.::
[;'-< 1500
metallurgy technigues, involves solidification. Castings obviously entail solidifica-
E-.."
tion; forgings and wrought products are also castings that have been hot worked,
and their behavior in many cases can be traced back to the method of solidification.
1000 In particular, the solidification rate of alloys is an extremely important processing
,
.......... variable. The solidification rate relates directly to the coarseness-or fineness-of
r - - - - - - - - - - - " - - - - - -.. dendritic structures and hence controls the spacing and distribution of micro-
segregates, such as coring, second phases, and inclusions. Thermal gradients
during freezing are also of great significance, being related to the formation of
t, sec microporosity in alloys. For these metallurgical reasons and from a process
engineering viewpoint, solidification heat transfer should be recognized as an
important topic.
The analysis of heat transfer during solidification is more complex than that of
conduction heat transfer presented in Chapter 9. This is one of the reasons for the
paucity of the literature devoted to this topic compared to that available for
conduction heat flow in solids. However, sufficient theory has evolved to treat many
practical problems, and the metallurgist should be aware of some of the analyses.
_l 329
::JU~u11uu1-.:aum1-ueat-i:nur.li1e.--- - -----~----- -----:a:u;--~~_cu;-:i:_----~--------------OUIIUll~-aUUll-iil~.U-D.U~IU~~-~
f ;
wher@ density of solidifying,metal, lbm/ft 3 , H 1 = latent heat of fusion of the
T, l-----r---TM
I metal, Btu/lbm, and M = thickness of metal solidified, ft.
I
I Equating Eq. (10.3) to Eq. (10.4) yields the rate at which the interface advances
I
Mold Solitl : Liquitl into the liquid: __ -~
L I
I dM = (Ii.t - T0 )fi{Z;,. , (10.5)
I
I dt p'H1 Jrr.t \
~~ ,I - -- ~----..____i
(oT)
Metal
Metal
qlx=O = -k - (10.2)
l l <01)
. r---~--":
~
OX x=O - J-:;;f; + +
Mold Mold
Remembenng that\ IX= k/pCP, \ve rewrite Eq. (10.2) as
\_ - ------ .'..L Mold Plane mold surface
qlx=O =
JkPCv
C ('li.t - To). (10.3) Convex mold surface Concave mold surface
v' nt
Fig. 10.2 Effect of contour on the heat flux into molds.
The product kpCv represents the ability of the mold to absorb heat at a certain rate
and is called the heat diffusivity.
The rate at which latent heat is evolved per unit area can be written As a first approximation, such effects are sometimes neglected because the
heated zone in the mold is shallow, and the difference in heat flow between a plane
,-- ---dA;r -- mold wall and a contoured wall is small. As such, we visualize that a given mold
p' Hr-d("' \ (10.4) surface area has t.he ability to absorb a certain amount of heat in a given time
' \
regardless of its contour. Thus, we generalize Eq. (I 0.3) for all contours, and for a For the infinite plate,
given surface area A, the mold absorbs an amount of heat Qin time t:
(IO.I I)
Q= J Aq/x=o dt = Ak(1Af - To) J.!!_
0
~ 0
Jr For the cylinder,
_ 2Ak(1Af - T0 ) r: (I 0.12)
- c .yt. (10.8)
.ynr:t.
For the sphere,
For a casting of volume V to completely solidify, all its latent heat must be
rt>moved; hence, the total latent heat Q evolved is
(10.9)
fJ = y(}n + 3~) ~ (I0.13)
Equations (10.8) and (I0.9) are then combined to yield the solidification time of a We can deduce Eq. (10.11) from Eq. (10.10) by a rearrangement. Equations
casting in terms of its volume-to-surface area ratio: (10.12) and (10.13) have resulted by rearranging expressions presented by Adams
and Taylor; 1 their expr~ion for the cylinder is approximate whily, that for the
-~
\ ,__::
"' \f (I0.10)
sphere is exact.
\\
\
where
n( 2
'
~--, l
p' H ) ( 1 ) I
'\ C. =4_ 1Af - JT0 .. . k[J,Cp
----~7~fo) . These expressions show the error of using Chvorinov's rule without regard to
p~P~~ ~\
the mold contours. For example, let us refer to Fig. 10.3 which relates freezing
y =\ p'HJ times for the three basic shapes according to Eqs. (I0.11)-(10.13). By calculating a
With these parameters, the' freezing times for the three basic shapes may be
compared. 1 C. M. Adams and H.F. Taylor, Trans. AFS 65, 170-176 (1957).
~------------------------------ ----------------
value of y from properties of the metal and mold, we read a value of fJ corresponding and for an infinite plate, VJ A = L where L is the semithickness. Therefore
~
to the different shapes from the curves of Fig. 10.3. For metal-sand combinations,
y is approximately unity, so that
t
= ____!!___ =
I.51 2r:t.
(2/12)2 ,
(1.51) 2(0.0178)
='~)
fJ (plate) = 1.13,
b) For a spherical casting, from Fig. 10.3,
fJ (cylinder) = 1.32,
V/A
fJ (sphere) = 1.38. fJ = r:: = 1.75,
y1r:t.t
We see that neglecting the contour can lead to an error of as much as 40-50 % in
freezing time. and for a sphere, V/A = R/3, in which R is the radius. Therefore
The expression for solidification of a sphere, Eq. (10.13), may be used for other R2 (2/12) 2
chunky shapes such as cubes with improvement in accuracy over the simple t = 9(1.752)r:t. = (9)(1.752)(0.0178) = 0.0564 hr.
relation, Eq. (10.11). Similarly, the expression for solidification time of a cylinder
may be used to approximate the freezing time of bars of square cross section. The sphere solidifies in less than one-tenth the time required for the slab to solidify.
Example 10.1 Determine the solidification time of the following iron castings, both
poured with no superheat into sand molds: 10.1.2 Effect of superheat on solidification time
""' .._r
a) a slab-shaped casting 4 in. thick; We can assess the effect of superheat on the solidification time by realizing that, in
b) a sp\h~\1::>'?\~1'1,\'~h
addition to absorbing latent heat, the sand must also absorb the superheat. Again,
er1ca y s ape d castmg
. 4 m.
. m. d"iameter.
we assume that the temperature gradients within the casting are negligible, and at
Iron data: the time solidification is complete, the entire casting is close to its freezing point.
In this case, the total quantity of heat to be removed from the casting is
Freezing temperature = 2802F,
-, Heat of fusion = 117 Btu/lbm,
' (10.13a)
- Solid density = 490 lbm/ft 3 ,
The subscript l denotes liquid phase properties, and.!\ T, is the amount of superheat
"" Liquid density = 460 lbm/ft 3 , in degrees. .
Heat capacity of liquid = 0.18 Btu/lbm F.
We now consider infinite plate castings, and in order to make the analysis
Sand data: simple, yet sufficiently accurate, we assume that Eq. (10.8) is valid even though the
interface temperature of the mold is not constant while the liquid phase loses its
Heat capacity = 0.28 Btu/lbm F,
superheat. In view of this approximation, it is certainly acceptable not to dis-
Thermal conductivity = 0.50 Btu/ft-hr F,
tinguish differences in the density of the liquid and solid phases. Thus p; ~ p',
Density = 100 lbm/ft. and when Eq. (10.13a) is set equal to Eq. (10.8), we obtain
Solution
n 1 ( p'H' )2( v)2 (10.14)
(Assume the mold)s initially at 82F:) t = 4 kpCP 'J/..t - IT0 A :
In this expression, Hj is the effective heat of fusion, and represents the sum of the
y =[ 2802 - 82]
(490)(117) (100)(0.28) = 1.33, latent heat of fusion and the_ liquid'~ supe_rh_t:a~,__t~Ljs,
. i Hj = l:f1-t.~;J,~TsJ (10.15)
Q50 2
2
r:t. = (100)(0.28) = 0.0178 ft /hr. Note that the solidification time is still proportional to (V/A) .
a) For plate castings, from Fig. 10.3, 10.2 SOLIDIFICATION IN METAL MOLDS
V/A When poured into metal molds, castings freeze rapidly, and temperatures change
fJ = r:: = 1.51,
y1r:t.t drastically in both the mold and the casting. An understanding of the variables
-- - a~~ting -,:,fic~t;on ;n metal mold' ;, impo<IBnt b~au'e mgo~l
mo'1
permanent mold castings, and all die castings are made in metal molds. Also many
-,--
t
sand castings are made in molds that incorporate metal inserts at strategic positions ------;;,-;;;.-:----Too =-------Too
/
to increase the rate of solidification.
~ The analysis of heat transfer when metal is poured against a chill wall is more
f'.<:r complicated than that when metal is poured into a sand mold; this is due to the
~ Iv
i &:. fact that metal molds are much better heat conductors than sand molds. The
J,f .added complexities are illustrated by the casting-mold situation ,shown in Fig.
10.4. At the solidified metal-mold interface, a temperature dr~p cle~ists, due to
thermal contact resistance. The condition of no contact resistance would exist
,/' ~-
only if the mold-casting contact were so intimate that wetting would occur, that
is, the casting would become soldered to the mold.
Mold ~Solid\
I
Liquid ----x (a) (b)
1
. - i -- - - - - - T M
Fig. 10.5 Analogous temperature distributions in (a) solidifying metal, and (b) a semi-infinite
solid.
.
field in the solidifying metal is between TM and T, rather than actuall_Y. exte~ding .to
T . However: the temperature "reaches" for T00 just as in the sem1-mfimte sohd,
a;d the temperature distribution within the solidified metal takes the form
To _ _ ___..- ' T- T
___s
x
= erf-- (10.16)
Too - 'I', 2fa
Fig. 10.4 The temperature distribution during the solidification of a metal from a chill wall.
In Fig. 10.5(a), T00 is not known a priori. It is an imaginary temperatu~~ whi~h
makes the temperature distribution analogous to the case of the sem1-mfimte
In addition to the contact resistance, there are other differences between
solid, or it may be thought of as an integration consta~t'. . .
solidification in sand and in chill molds:
We now develop an expression for the rate of sohd1ficatlon by applymg the
a) The thermal conductivity of the metal being cast forms an important portion
boundC;lry conditions
of the overall resistance to heat flow. This results in the surface temperature being
well below the melting point, while appreciable thermal gradients exist within the T(M, t) = 1iJ, (10.J7)
solidifying metal.
b) More total heat is removed during solidification because of the solidified metal i.e., the temperature at the solid-liquid interface is the freezing point. In addition,
which is subcooled. Thus, the heat capacity of the solidifying metal is important. we "'recognize that the rate of evolution of latent heat of fusion equals the heat flux
In the following sections, we shall discuss selected problems of solidification, into the solid at the interface, that is,
each of them being of some practical importance.
,_ oT dM
10.2.1 Constant'ca~tmg surface temperature k'-(M t) =Hp'-
OX ' f dt
(10.18).
,,.,.. / . ?
Consider a mass of pure liquid metal, initially at its freezing temperature, which
has its surface suddenly cooled to a constant temperature T,. After sofne solidifica- When applied to the temperature distribution, Eq. (10.17) yields
tion has occurred, the temperature profile in the solidifying metal will appear as the
solid profile (Fig. IO.Sa). The temperature profile is identical to the temperature 1i.t - T, M
(10.19)
---=erf--
profile in the semi-infinite solid depicted in Fig. I 0.5(b ), except that the temperature Too - T, 2fa
I
------------- - - - - - - - - - - - - - - - -- - - - - - - _ _ , . . . . , . - - - - - - - - - - - -
'
Since the left-hand side of this equation is constant, the argument of the error (
function must also be constant. Hence il 1.2
M = 2f3fo. (10.20)
Scale changes
Once again the thickness solidified is proportional to r". 0.8
To evaluate the constant /3, we evaluate the heat flux at the solid-liquid
interface which we obtain from Eq. (10.16):
0.4
k' oT (M, t) =(Too - T,)~ exp [- M1] ...
ox JnJt 4a't ...
0.3
= (TM - T.)~ exp ( - [3 2 ). (10.21) ...
Jnjterf P
The latent heat evolved at the interface is written
JP -;z;-
H'dM_H'R
JPv . (10.22)
0
Substituting Eqs. (10.21) and (10.22) into Eq. (10.18), and simplifying, we have
( T - T) _ P _ = {Je~' erf il
We now have an expression to calculate f3. We may use Fig. 10.6 to evaluate /3, M 'Ht../11
rather than using Eq. (10.23) which entails trial and error. Fig. 10.6 Evaluation of f3 for Eq. (10.23).
To summarize, f3 can be determined from Fig. 10.6. Thus the rate of solidifica-
tion is known (Eq. 10.20), T00 can be determined (Eq. 10.19), and the temperature
distribution can be computed (Eq. 10.16), if so desired.
This analysis is, of course, valid for unidirectional heat flow; its results can be 2.4
R'
applied to slab-shaped castings. If the solidification time is sought, then evaluate f3 I
and use Eq. (10.20) with M = L, the semithickness of the slab. ::;: ::i:::"
t:, 2.0
The above method of solution is not systematic and cannot be extended to the (.)"
solidification of other shapes. However, Adams 2 has presented a method utilizing
a power series that can be extended to the freezing of spheres and cylinders. His
results for solidification times of spheres and cylinders freezing with a constant
surface temperature, T,, are given in Figs. 10.7 and 10.8.
While application of the foregoing is limited because it is difficult to imagine .. 0.8
practical situations in which a constant surface temperature is maintained, an
example of a case of practical importance is the determination of the solidification
rate in a large steel ingot poured against a copper, water-cooled mold wall, except
for the initial stage of solidification. The solutions are also useful for indicating the
8
maximum freezing rate that can possibly be obtained by convective cooling, since
R 2 /or.'t
the boundary condition of constant surface temperature corresponds to a case of
h --> oo at the surface.
2 Fig. 10.7 Solidification times for spheres with constant surface temperatu~e: (F~om C. M.
C. M. Adams, "Thermal Considerations in Freezing," Liquid Metals and Solidification, t Adams, "Thermal Considerations in Freezing," Liquid Metals and Sohdificatwn, ASM,
ASM, Cleveland, Ohio, 1958. Cleveland, Ohio, 1958.)
,
M--J
I
\-.--(TM-To)(~) l and
I
x1 i
x=O and Fig. 10.10, which wa~ d~~i~~d by calculating f3 on a trial and error solution
Fig. 10.9 Temperature distribution during solidification with no interface resistance. of Eq. (10.26), and then determining Ts from Eq. (10.23).
I
----------~-------
2 .
p'C~
k'
If we wish to calculate the thickness of the solidified metal, then we determine
' T, from Fig. 10.10 and use Fig. 10.6 for a value of f3. With this value of /3, Eq. (10.20) 2
I,
can be applied. = (2/12) (490)(0.16) = 0.0118 hr.
(4)(0.98)2 (48)
Example 10.2 Determine the freezing time of an iron slab-shaped casting which is C'
4 in. thick. _Assume no interface resistance, and consider the cases of pure iron c) (TM - T0 ) _l1_ = 3.72, as in part (a);
being poured at its freezing point into (a) a sand mold, (b) a water-cooled copper HI
mold, and (c) a very thick copper mold. (48)(490)(0.16) = 0.570.
Mold data: (230)(560)(0.09)
Heat capacity, Density, Thermal conductivity, From Fig. 10.10, we get
Material Btu/lbm F lbm/ft 3 Btu/ft-hr F
T, - To ~ 0.63,
Sand 0.28 100 0.50 TM - To
Copper 0.09 560 230 T, = (0.63)(2802 - 80) + 80 = l 790F.
L
-----~~-----
Then we resort to Fig. 10.6 to obtain a value of f3: As with sand castings, if the temperature gradients are negligible, then T, ~ TM,
and only latent heat is removed from the casting during solidification. Therefore
(TM - T,J Jn
C'P!: = (2802 - 1790) ( 0.16 ) = 0.78 by combining Eqs. (10.9) and (10.29), we can easily show that
H1 .y n 117 n V h(TM - To)
- = t. (10.30)
and A p'H1
Note that shape has no effect on the applicability ofEq. (10.30). Shape was also not
/3 = 0.70. specified in the case of the cooling or heating of a solid body with negligible internal
Using Eq. (10.20) as in part (b) yields temperature gradients. Ifwe wish to apply Eq. (10.30) to unidirectional solidifica-
tion, we see that
t = 0.0232 hr.
h(TM - To)
1-~ M = 'H t. (10.31)
10.2.3 Interface resistance p J
In the previous sections, we depicted situations in which the surface temperature The thickness solidified, M, is proportional to time rather than the square root of
T, remained constant. When we examine the more likely situation of some inter- time.
face resistance, then the surface temperature of the casting varies with time, and Now consider the case in which T, =f TM, and the heat leaves the casting via h
the analysis becomes more complex. Before analyzing the temperature history in at the surface to a. water-cooled mold maintained at T0 . Here we simplify the
the mold and casting in a general case (Fig. 10.4), let us first consider a simpler case analysis by approximating the temperature profile within the solidifying metal as a
(Fig. 10.11 ). linear function. Then we write the heat flux at the casting-mold interface as
'TM - T, (10.32)
Mold Solid Liquid qlx=O =k M
Also
qlx= o = h(Ts - To). (10.33)
We then eliminate the surface temperature T,, which varies, by combining Eqs.
(10.32) and (10.33); because of the linear temperature profile, we express the flux
at the solid-liquid interface simply as
To _ _ __, I
I TM - To
M-iI qlx=O = qlx=M = l/h + M/k' (10.34)
I
I
I
In addition, at x = M, the latent heat is evolved so that
I
dM
Fig. 10.ll Temperature distribution during solidification with a predominating interface qlx=M = p'HrJ.t (10.35)
resistance.
By combining Eqs. (10.34) and (10.35) and integrating with M = 0 at t = 0, and
In Fig. 10.11, the interface resistance predominates over the resistances offered M-= Matt= t, we obtain
by both the solidifying metal and the metal. Practical importance is attached to M = h(TM - To) t - _.!!_ Mz. (10.36)
this case when solidification time is short; the analysis is u-seful for estimating p'H1 2k'
solidification times of small thin-section parts cast in heavy metal molds such as Adams 3 has solved the problem in a more rigorous manner in which the tempera-
die and permanent mold castings. ture profile is not assumed to be linear as above. The more refined analysis is
In this case, the temperature gradients within the mold and the casting are similar to Eq. (10.36) with an additional factor a:
negligible, and the heat escapes the casting as if a heat-transfer coefficient applies
at the surface. Thus the total quantity of heat Q that crosses the mold-casting = h(TM - T0 ) t _ _.!!_ Mz. (10.37)
M p'H a . 2k'
interface in time t is 1
3
Q = hA(T, - T0 )t. (10.29) C. M. Adams, ibid.
!
_J_
......--. ..----
-- - - - - - - - - ..
~------------ ~onomcauutrI11-inenn--.no1os-~
1 C~(TM - To)
4+ 3H
f
. ;--------TM
Equation (10.37) is almost exact for hM/k > 1/2. For hM/k < 1/2, the thickness
solidified is overestimated by approximately 10-15 %.
10.2.4 Gradients within mold and casting with interface resistance
In tackling this situation, we extend the concept of the mold-metal interface
temperature introduced in Section 10.2.2. However, in this case there is no constant
. "' temperature at the casting-mold interface; rather, there are two surface tempera- Imaginary
plane at - - - - -
tures, neither of which is constant. To handle this problem, consider an imaginary T,
reference plane between the mold and casting which is at T,, where T, is constant
and determined by the method of Section 10.2.2 using Eq. (10.28). The contact
resistance is then apportioned on both sides of the imaginary plane in accordance M
with the equations
To
(10.38)
Fig. 10.12 Temperature distribution in the casting and mold when the interface resistance
does not predominate.
(10.39) M
use Eq. (9.113). For the mold side, heat is supplied to the mold from a source 2hc /
at T, to.the surface at T,M via hM.
We present Fig. 10.13 to illustrate the effect of the contact thermal resistance.
First, in order to solidify the same amount of metal, it takes more time with some Fig. 10.13 Rate of solidification in metal molds with and without contact resistance.
resistance. Note, also, that the two curves become parallel after the early stages of
solidification, and M varies linearly with .jt for both cases. For the early stage of
in Example 10.2 when it is cast in a very thick copper mold. The total heat transfer-
solidification, M does not vary linearly with .jt when there is some contact
coefficient across the casting-mold interface may be taken as 250 Btu/hr-ft 2 F.
0
resistance.
b) When the casting has completely solidified, what is the surface temperature of
Example 10.3 a) Determine the freezing time of the 4 in. thick slab of iron discussed the mold?
Solution From Fig. 9.23, we have
a) First solve the problem as if there were no interface resistance, and obtain T,. _T_-_T..;_
0
~0.86,
This was done in Example 10.2 with the result T, - To
so that
T, = 1790F. T = (0.86) (1790 - 80) + 80
Using Eq. (10.39), determine the coefficient for the casting side of the interface: = 1550F.
Up to this point, we have recognized that a thermal resistance exists at the
(48)(490)(0.16)) mold-casting interface in all cases unless soldering occurs. Since no physical
he = (1 + (230)(560)(0.09) (25 0) bonding takes place at the interface, then the casting and the mold are free to move
due to thermal-physical effects. In fact, it is well known that when a metal is cast
~. = 392 Btu/hr-ft 2 F.
against a metal mold, a gap forms at the interface. This gap is formed as a result of
Now use Eq. (10.37), but in a modified form for the case at hand; it is the mold expanding due to its absorption of heat, and to the solid skin of metal
shrinking due to its lowering temperature. Usually, the prediction of the rate of heat
M = hc(TM - T,) t _ he M 2.
transfer across the gap is not reliable, and so we rely upon empirical measurements
p'Hfa 2k' of surface temperatures and the heat absorbed by the mold to deduce appropriate
Substituting properties and M = i ft, we have heat-transfer coefficients. Some heat-transfer coefficients for various casting situa-
tions are given in Table 10.1. Since such values are strongly dependent on specific
~ = (392)(2802 - 1790) t - (392) (~)
2
processes and geometrical situations, the data in Table 10.1 should serve only as
6 (490)(117)a (2)(48) 6 ' general guidelines.
in which Table 10.1 Heat-transfer coefficients across casting-mold interfaces
1 1 (0.16)(2802 - 1790) Heat-transfer
a=-+ 4+ (3)(117) = 1.34. coefficient, Reference
2
Casting situation Btu/hr-ft2 F listed below
Solving for t yields
Steel in continuous casting moid 50-400* (a)
t = 0.0543 hr.
Steel in continuous casting mold, 4 x 4 in. (b)
b) To solve for the surface temperature of the mold we may use Eq. (9.113) or withdrawal rate of 20 in./min 85
Fig. 9.23) with h replaced by hM, TI replaced by T,, and T; replaced by T0 . withdrawal rate of l 00 in./min 140
withdrawal rate of 175 in./min 190
For the surface,
Ductile iron In gray iron mold (coated with
x/(2j;i) = 0. amorphous carbon) 300 (c)
10.3 INTEGRAL SOLUTION FOR SOLIDIFICATION temperature below the freezing point. 4 This is the same problem as we considered
In Chapter 7 we used an integral method to examine a problem involving heat in Section 10.2.1.
transfer with convection, in which all the thermal gradients existed within a thermal We apply Eq. (10.44) to the freezing problem at hand by taking x' as the thick-
ness of material solidified, M(t). The conditions to be met are
boundary layer. A thermally affected zone in conduction problems can be examined
in a similar manner. The temperature transient within a body is caused by some M(O) = 0, (10.45)
boundary condition at the surface (x = 0). The thermally affected zone is defined
between x = 0 and x', an interior position at which the body has just begun to feel
T(O, t) = T,, (10.46a)
the transient introduced by the boundary condition. As long as the system is in an T(M, t) =TM, (10.46b)
unsteady state, x' is a function of time.
and at the liquid-solid interface,
Before proceeding to solve a specific solidification problem, we first develop
'i, the integral equation of conduction heat transfer. Consider the differential equation , oT , dM
that applies to one-dimensional conduction heat transfer with constant thermal k ox (M, t) = p H1 dt. (10.47)
diffusivity, namely,
Let 8 = T - TM; then the boundary conditions transform into
o2 T oT
CJ.--=-
2 (10.40) 8(0, t) = T. - TM = 8,, (I0.48a)
o x ot
8(M, t) = 0, (10.48b)
We multiply both sides of Eq. (10.40) by dx, and integrate from x = 0 to x': and
x' x'
f o T
2
a ox 2 dx =
f at
oT
dx. (10.41)
08
0
dM
k' )M,t) = p'H1 dt (10.49)
0 0
Expressing the integral equation in 8, and applying Eqs. (10.48a), (10.48b), and
We' readily visualize the left-hand side of Eq. (10.41): (10.49) yields
x' M
f
2
We rely on a mathematical identity to represent the right-hand side ofEq. (10.41): As usual, when employing an integral technique, we assume the temperature
x' x' distribution. In this case, let us assume that the temperature distribution is a
JT dx - second-degree polynomial in x. Then
f at
0
oT
dx =
d
dt
0
dx'
T(x', t) dt. (10.43)
8 = a(x - M) + b(x - M) 2 , (10.51)
which automatically satisfies Eq. (10.48b). Two further boundary conditions must
Substituting Eqs. (10.42) and (10.43) into Eq. (10.41) yields the conduction heat be sadsfied: Eqs. (10.48a) and (10.49). However, Eq. (10.49) in its present form is
transfer integral equation not suitable because the coefficients in the polynomial would involve dM/dt. In
x'
turn, the left-hand side ofEq. (10.50) would involve dM /dt, and the integral equation
oT , oT
CJ. [ ox (x, t) - ox (0, t)
J= dtd J Tdx - dx'
T(x', t)dt (10.44)
would then yield a second-order differential equation for M(t), whereas there is
only one initial condition for M(t), namely, M(O) = 0. To overcome this difficulty,
0 we transform Eq. (10.49) into a different form before applying it to the assumed
polynomial.
Several problems involving phase changes have been solved by the integral
method which we can use to obtain approximate solutions. To apply this method, 4
The remainder of this section is from T. R. Goodman, "Integral Methods for Nonlinear
consider solidification when the surface of an ingot is maintained at some constant Heat Transfer," Advances in Heat Transfer 1, Academic Press, New York, 1964.
From Eq. (10.48), we write
and
b = aM +es, . (10.56)
M2
Wsk'
=
Substituting the completely described profile into Eq. (10.50), we obtain the
Pinch (supporting rolls)
differential equation for M: .
MdM = 6a'[l - (1 + )112 + ]_
(10.57)
dt 5 + + (1 + )1/2
Integrating Eq. (10.50), with the initial condition M(O) = 0, leads to Fig. 10.14 A strand casting machine.
M = 2f3fo. (10.58)
Here As the metal passes through the mold, we assume that a constant heat-transfer
- 7Y>~--o->y.,~,-=-(\1-~~~-~~-~~~---
\, ~~,
~
1'..
d= hM ' 2.0
k'
1.8
I
I
I 1.6
I
1.4
1.2
1.0
0.8
0.6
x=O
x-
Fig. 10.15 The partially frozen portion of a continuous casting and the temperature profile as
,a function of the distance down the mold.
0 or~ . 0.3
i~-z I
0.4 0.5 0.6 0.7 0.8 0.9
h2y
uk'p'C~
1.0
Fig. 10.16 Thickness solidified, M, versus distance down the mold. (Figures 10.16-10.18 are
from A. W. D. Hills and M. R. Moore, ibid.)
liquid superheat is conducted toward the mold wall through the solid skin, and in
the formulation the effective latent heat of fusion Hj applies in which
Hj = H1 + cp,1(Tp - TM).
"
The results are convenient when presented in dimensionless groups. From
Eq. (10.37), we deduce the groups
1)
h2y
---, hlh"
I I
1.0
uk'p'c; h" ~ 0.9
~I(
c;(TM - To)
2) '
l~ 0.8
H'J
0.7
hM
3)
k' 0.6
These groups and others are used in Figs. 10.16-10.18 to present the results. We 0.5
can determine the thickness of solid metal at various positions down the mold from
Fig. 10.16. Figure 10.17 shows how the surface temperature T, of the cast metal
.!. 04
0 0.1 0.2 0.3 0 .4 0.5 0 .6 0. 7 0.8
~ h2y
varies with the mold position, and we use Fig. 10.18 to calculate the total rate of
uk'p'C~
heat removed entering the mold oflength L. Skin thickness and surface temperature
are important parameters, since the solid skin on exit from the mold must be of
some minimum strength. The rate of heat removal is useful to compute cooling Fig. 10.17 Surface temperature versus distance down the mold. Numbers on the curves are
water requirements. the same as in Fig. 10.16.
r
--- --- ------ - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - _ , _ , , - - . . . - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
~
v~- ~
~ 0.7 h2y
-l
~klI ~ 0.11.
u Pc~
~ 0_6
~ -----------
Therefore
= L = (~l) (120/12 x 60)(44J(4(o)(O~),
y - (250) 2
t = 3.6-ft,
r;JC,L
~
r10l-yv
b) Frnm F;g. I0.18, _./"" \)EQIM,;:11W J
10.1 Plot distance solidified versus the square root of time for the following metals (in each
case the pure metal is poured at its melting point against a flat mold wall):
and a) Iron in a green sand mold.
H'f 131 / - b) Aluminum in a green sand mold.
/ = 0.308. c) Iron in a mullite mold heated to 1800F.
0.16(2730 ~j
I
------------------------------------~,.
10.2 How long does it take to freeze a 4-in. diameter sphere of pure iron in a green sand mold 10.6 A 2-in. thick slab of aluminum is cast in a mold made of sand (forming one face) and a
assuming proprietary material (forming the other face). The aluminum is poured with no superheat, and
a) no superheat and neglecting the heat flow divergence? the as-cast structure of the slab is examined after solidification and cooling. The examination
b) no superheat and realizing that a sphere is being cast? shows that the plane of last solidification (i.e, the plane where the two solidification fronts meet)
c) 200F superheat and realizing that a sphere is being cast? is located I! in. from the sand side. Knowing this and using the following data, calculate the
heat diffusivity (not the thermal diffusivity) of the proprietary material.
10.3 Plot distance solidified versus time for iron poured at its melting point into a heavy copper
mold, assuming that
a) there is a large flat mold wall and no resistance exists to heat flow at the mold-metal
interface;
b) interface resistance to heat flow predominates (h = JOO Btu/hr-ft 2 F);
~' c) interface resistance predominates in the early stages of solidification and then becomes
negligible.
10.4 Show whether or not iron can be cast against a very thick aluminum mold wall without
causing the aluminum to melt.
10.5 Slab-shaped steel castings are prone to center-line porosity, which-for our purposes-
is simply an alignment of defects along the plane of last solidification. The sketch below shows
the solidification of a slab cast in sand and the location of the defects (centerline porosity).
Data:
Freezing point of Al= 1220F
Al Sand
10.7 A continuous casting machine forms molten steel into a slab up to 76 in. wide and 9 in.
The solidification time for the 2-in. slab cast in sand is known to be 6 min; when cast in an
thick at a production rate of 1.5 million tons per year. Determine the vertical length of the
insulating mullite mold, the time is 60 min. Ifa casting is made in the composite mold depicted
mold if the solid shell must be 0.5 in. thick at the mold exit, and calculate the cooling water
below, determine the thickness T the casting should have to yield li in. of sound metal after
machining. requirement (lbm/hr) if its temperature rise is from 80F to 180F.
TM= 2770F,
p_= 480lbm/fi3.
.l
10.8 The dwell time in the mold of a continuous casting machine is defined as the interval
which the metal spends in the mold during solidification, that is, t = L/u, in which t = dwell
time, L = length of mold over which solidification is occurring, and u = velocity of metal
through the mold. Since the skin solidified in the mold is not thick, a simple analysis might be
expected to apply. 11
a) Assume that the temperature distribution within the solidified metal is linear at any
position down the mold and show that
RADIATION HEAT TRANSFER
b) From part (a) obtain an expression for thickness solidified versus time.
c) Compare the results for dwell time in Problem 10.7 and the dwell time calculated from
part (b) for the same conditions.
In Chapter 6, we alluded to energy transfer by radiation as a mechanism for heat
10.9 A junction-shaped casting as depicted below is made in a sand mold. The junction may
flow, but due to the entirely different natures of conduction and radiation, only a
be considered infinitely long in the z-direction. For the upper right quadrant of sand.:
limited discussion of radiation appeared prior to this chapter. Energy transport
by conduction depends on the existence of a conducting material. On the other
hand, radiation is electromagnetic energy in transport; therefore, energy travels
through empty space via radiation. The rate equations for conduction and radia-
-,. tion reflect their very different characters, and show no similarity. The energy
flux by conduction is proportional to the thermal gradient, that is, qx = -k(oT/ox).
In radiant heat transfer, the energy flux-the so-called emissive power-is propor-
tional to the fourth power of the absolute temperature, that is, e oc T 4 .
Solids, liquids, and some gases emit radiant energy composed of many wave-
lengths. Figure 11.1 depicts the spectrum of electromagnetic radiation and the
atomic mechanisms responsible for the various forms of radiation. The excitation
process causing radiation may vary and can be classified as fluorescence,
x-radiation, radio waves, etc. If the excitation comes from molecular motion which
a) Write a differential equation for temperature. characterizes temperature, the radiation is called thermal. The thermal radiation
b) Write the boundary conditions (for time and space) that apply. is of primary interest here and its range of wavelengths is taken to be 0.1:-100 .
c) Write the solution yielding temperature as a function of position in the sand and time.
11.1 BASIC CHARACTERISTICS
j 361
~--------- .. - - - - -------- -- ------- ---- ---. --.-r,~------------
17 log 10 ;\ (with;\ in A) reflected), and r = transmissivity (fraction of G transmitted). Most liquids and
16 solids are opaque to thermal radiation, so that r = 0, and
15
a+p=l. (11.2)
14
1
I km- 13
11.2 THE BLACK RADIATOR AND EMISSIVITY
12
. Electrical conductor
11 Radio waves----- carrying alternating In the study of real surfaces, it is convenient to define a hypothetical ideal surface
I m~ current
called a black radiator. A black radiator is a surface which absorbs all incoming
radiation; thus a = 1, and p = 0. Another important characteristic of the black
80
I cm radiator is that it is a perfect emitter, since it emits radiation of all wavelengths, and
I mm 7
its total emissive power is theoretically the highest that can be achieved at a given
----r---- 6 temperature. It follows that any real body emitting thermal radiation has a total
Thermal 5 ~~~ni~;;:~e~=-~-= ~:::~~:: :~~:~~~:~s emissive power of some fraction of the emissive power of a black radiator, which is
radiation I 4 t t V1s1ble often called a black body. This fraction is defined as the emissivity e. Therefore
Displacement of outer
----L--- _ 3 - - -} electrons of an atom (11.3)
2 traviolet
I m - I x-rays----- Displacement of inner Imagine a region in a space completely filled with black body radiation. A
I A_ o ~ l electrons of an atom
real body 1 emitting radiation at a rate of e 1 is placed in this region; the net rate
-1-f--I- '>-of energy transferred from the body is
-2 'Y-rays - - - - - - Displacement of nucleons
.i. in an atomic nucleus (11.4)
-3-'
Fig. 11.1 Electromagnetic radiation spectrum (I is 10- 3 mm; Im is 10-3 ).
energy energy
emitted absorbed
~
J=e+pG If the body is in thermal equilibrium with the black body radiation, then qnet = 0,
,..---"-.,
and a 1 = e 1 . The same result can be obtained for any other body placed in this
tpG
space, and thus the absorptivity and emissivity of any body at thermal equilibrium
-----~------------ are equal.
' TG
Fig. 11.2 The distribution of thermal radiation at a surface.
Electric
heaters
ene_rgy leaving the surface of a body. The radiosity includes both the energy
~mitted and the energy reflected by the surface. As fluxes, the units of G and J are
m Btu/hr"ft 2
. The incoming energy (total irradiation) can be absorbed, reflected, or trans-
mitted, so that . (b)
walls at a uniform temperature. Since the hole is small, only a very small fraction 12
of the radiation entering the hole and diffusely reflected off the nonblack walls will ~1N~
~!
escape; hence the escaping energy is entirely representative of the radiation within .c JO
the cavity. _,;
"'x
The energy emitted from any point on the walls of the enclosure is seb; when 8 '-'
~
~
I
this radiation strikes other parts of the enclosure, a certain fraction is reflected, s
namely, pseb. Similarly, a second reflection equals p(pseb), and the third one is
>
6 :_i
3 11
p seb. Since the hole is small, probability favors radiation that ultimately passes 11
11
through the hole to be made up of an infinite number of such reflections. Therefore 4
and since a = 1 - p, and thermal equilibrium exists within the cavity, then Fig. 11.4 Black body spectral energy distribution.
s = 1 - p for the isothermal cavity. Therefore Eq. (11.7) reduces to
Substituting Eq. (11.10) into Eq. (11.9) and integrating result in the Stefan-Boltzmann
e (hole)= eb. (11.8) equation for black body radiation, which is often written as:
Since radiation from a cavity is very nearly black, the hohlraum (heated cavity), "- eb = <JT4, (11.11)
shown in Fig. l 1.3(b), is normally used as a radiation standard. Such an enclosure
may be constructed with heavy metal internal walls that are roughened and oxidized where the constant <J (in engineering units) is
to provide a surface of high emissive power. The walls are heat~d electrically, and 2n 5 K~
the exterior is well insulated. <J = - - - = 0.1713 x 10- 8 Btu/hr-ft 20 R 4 . (11.12)
15c 2 h3
11.3 THE ENERGY DISTRIBUTION AND THE EMISSIVE POWER In view of the Stefan-Boltzmann equation, the total emissive power e of a real
~ body is
The thermal radiation from a solid body is composed of a continuous spectrum of
wavelengths forming an energy distribution. Figure 11.4 depicts the spectrum of a (11.13)
black body at various temperatures. For most engineering calculations, the total According to Fig. 11.5, the ratio e/eb of a typical real body varies with wavelength;
emissive power of a black body at a particular temperature is an important quantity, thus the quantity s used in Eq. (11.13) is in fact the mean value of a,_, the mono-
which is given by the area under the curve applied at that temperature. Therefore chromatic emissivity. Qualitatively, the monochromatic emissivity for metals
00 decreases with increasing wavelength. In contrast, the monochromatic emissivity
eb = J
0
eb,). dA, (11.9)
T= constant
e,
where A is the wavelength, and eb ,_is the monochromatic emissive power. Planck's
equation or Planck's distribution l~w relates eb,'- erg/cm 3 sec to the wavelength (cm) Planck's law (black body)
and the absolute temperature:
A typical real surface
2nhc 2 ;_- 5
eb,'- = exp(~) - 1,
(11.10)
KBAT
where h = Planck's constant, 6.6238 x 10- 27 erg-sec, c = velocity of light, X
10
2.9979 x 10 cm/sec, and K 8 = Boltzmann's constant, 1.3803 x 10- 16 ergjK. Fig. 11.5 Comparison of black, gray, and real surfaces.
----~---------------------
of electric nonconductors has a tendency to increase with increasing wavelength, Figures 11.8 and 11.9 give the em1ss1v1ty of several substances including
but the variation of e, with A. can be quite irregular. Figure 11.6 depicts experi- polished metals. Additional data are given in Table 11.1; several oxides and some
mental results showing these trends for aluminum having three different surface refractory materials are listed. More extensive listings of emissivity data are
finishes. For the bare metal surface (commercial finish), the emissivity decreases available in texts on radiation heat transfer. 1 2 3 In metals, emissivity increases
almost steadily with increasing wavelength. The anodized aluminum behaves as a with electrical resistance, and, correspondingly, with temperature; E is nearly
nonmetal, since the anodizing process produces a thick* oxide coating on the proportional to T 112 for many pure metals. For electric nonconductors, the
surface. Furthermore, by comparing the commercial finish surface with the emissivities are much higher than for Clean metal surfaces, and they often, but not
polished surface, we see that roughness increases emissivity. But since the descrip- always, decrease with temperature.
tions such as polished, rough, or commercial finish are rather vague, emissivity
:::- 1.0
data should be used judiciously with regard to the surface description.
Not only does emissivity vary with wavelength, but in the case of well-polished :~
surfaces, it is also a function of the view angle from the normal to the surface
e 0.9
UJ
(Fig. 11. 7). Therefore, a directional emissivity Eq, exists, but for our purposes the E
used for calculations is the average value for a substance emitting radiation in all
0.8 ~ough Beryllium oxide.
directions as well as all wavelengths. In this sense, E used for most calculations is black disk
properly called the total hemispherical emissivity; usually, however, it is simply
0.7
called emissivity.
;; 1.0 BraS>. oxidized
0
.;; 0.6 ---
0.8
Aluminum, 0.5
anodized
0.6
0.4
0.4
0.3
0.2
Aluminum.
0.2
Aluminum alloy.
Wavelength A, 0.1
775 ST,
Fig. 11.6 Typical wavelength dependence of s, and a, for metals. polished
~ 1.0 Brass. poli>hed
0
>,
Al 2 0 3
~0-----=5700::--~10~0~0,-----1~5~00.,-----2~0~0~0-~2~5~0~0--30~0Lo_ _3~500
;;:
-~
0.8 Temperature. F
a Copper oxide (a)
" 0.6
--;
c Fig. 11.8 The emissivity of several materials. (From G. G. Gubareff, J. E. Janssen, and
0
u R.H. Torborg, Thermal Radiation Properties Survey, Honeywell Research Center, Minneapolis,
0.4
:!:! 1960.)
i5 Cr
0.2
1
Al H. C. Hottel and A. F. Sarofim, Radiative Transfer, McGraw-Hill, New York, 1967.
0 2
0 45 90 E. M. Sparrow and R. D. Cess, Radiation Heat Transfer, Brooks/Cole, Belmont, California,
Fig. 11.7 Variation of directional emissivity. The surfaces are well polished. 1966.
------ 3
* Relative to the depth of participating material which is only a fraction of a micron for metals. T. J. Love, Radiative Heat Transfer, Merrill, Columbus, Ohio, 1968.
---------;r-------~r-----,,
.'.:' 1.0
-~
~
w 0.9 .'.:' 1.0
;;: Iron, electrolytic, oxidized
~
0.8 w 0.9
Cast iron,
0.7 oxidized
0.6
0.5 0.6
0.5
0.3 0.4
Iron, wrought,
0.3 smooth
~st iron,
polished
Iron, pure,
0.2 polished
Gold,
~===----------polished
0 0.1
0 500 1000 1500 2000 2500 3000 3500
Temperature, F
{b)
400 800 1200 1600 2000 2400 2800
Fig. 11.8 (continued) Temperature, F
'
Table 11.1 The emissivity of some ceramic materials. (From H. C. Hottel and A. F. Sarofim, Fig.11.9 Variation of emissivity with temperature and condition of ferrous materials. (From
ibid.) G. G. Gubareff et al., ibid.)
I
-------~-xt...~ange-uerween-mnnne-parauerp1aTe~TI
in which T* = ft(I';. We illustrate the use of these approximations in the Fig. 11.11 Radiation exchange between parallel surfaces. (a) Irradiation from surface (2).
following section. (b) Irradiation from surface(!).
~ 0.7
because the whole amount of the radiation leaving one of the surfaces is completely
>
:g intercepted by the other surface. We shall treat more complex geometrical aspects
~ 0.6 in the next section, but here we consider a case of simple geometry to show the
utilization of the system properties. Specifically, consider the two parallel surfaces
0.5
of infinite extent in Fig. 11.11. In this schematic representation, the energy is
viewed as two parts: the radiation originating from surface 1 is shown in Fig.
11.1 l(b); radiation originating from surface 2 is shown in Fig. 11.1 l(a).
0.4 The irradiation of surface 1 from surface 2 consists of all the upward pointing
arrows in Fig. 11.1 l(a), or
0.3
ez + P12P22e2 + Pi2P~2e2 + P~2P~2ez. (11.15)
We write this infinite series as
0.21------
Magnesium oxide
(11.16)
0.1 1- P12P22
In a similar manner, the summation of the upward pointing arrows in Fig. 11.1 l(b)
3500
is
500 1000 1500 2000 2500 3000
Temperature. F
Fig. 11.10 Two materials that exhibit a range of constant i; and a range of rapidly varying e. (11.17)
1- P11P21
11.5 EXCHANGE BETWEEN INFINITE PARALLEL PLATES The sum of these two terms is the total irradiation of surface 1 ; thus
When a system is of simple geometry such as infinite parallel plates, concentric ez P21e1
~~~~-+ ~~~~- (11.18)
cylinders, or concentric spheres, the evaluation of radiant heat transfer is simplified 1 - P12P22 1- P11P21
I
r
,..,----------------------------- -----~ ------------------------------ - - - - - - . - . = - - - . - - . . . . - - - - - - - - - - - - - - - - - - - - - - e - - - - -
The net heat fluJ<-from surface 1 is Example 11.1 We can use the above results to illustrate the principle of radiation
shields. Multiple shields are very effective when used as high-temperature
(11.19) insulation against thermal radiation. To illustrate the principle, develop an
but the radiosity 1 1 is composed of the radiation emitted at surface 1 plus that expression for the reduction in radiant heat transfer between two infinite and
reflected, parallel refractory walls, when a thin plate of aluminum is placed between them..
Assume that all surfaces are gray; the t'.missivity of the walls is 0.8 and the emissivity
(11.20) of the aluminum is 0.2.
(The secon~ subscript of pis not denoted at this point.) By combining Eqs. (11.18)- The system with appropriate subscripts is shown below.
(11.20), we find that the net heat flux from surface 1 is
Aluminum
I plate
or
tX12e2 tX11P21e1 )
ql,net =el - ( + . (11.21) Refractory Refractory
1- P12P22 1 - P11P21 wall wall
Here a 1 x is denoted a 12 and a 11 for the first and second terms on the right side of
Eq. (11.21) because it is associated with the radiations originating at surfaces 2 and
1, respectively.
The basic practical difficulty in the analysis of systems with nonblack surfaces
is what absorptivity value should be used. In the light of this problem, which was
discussed in Section 11.4, let us analyze the various forms that Eq. (11.21) takes Solution. First, we examine the case in the absence of the aluminum sheet, that is,
depending on the manner in which we evaluate a 11 . The gray assumption which direct radiant transfer from surface 1 to surface 3. Equation (11.22) applies with
appropriate change of subscripts. Therefore
would apply for Inconel-X plates, if both were between 150 and 1000F, and for
refractory surfaces at high temperatures such as magnesia in Fig. 11.10 for tempera- q 1,neiCno shield) = (eb 1 - eb 3 ) - - - - - -
tures greater than 1500F, makes a 12 = a 11 = i; 1, a 21 = a 22 = i; 2, and p 12 = l/e1 +
1/e 3 - 1
P11 = 1 - e 1, p 21 = p 22 = 1 - e2. In this case, Eq. (11.21) simplifies to Now in the presence of the shield, we apply Eq. (11.22) to the net radiant flux
1 froni surface 1 to surface 2. Hence
q -(e -e ) - - - - - (11.22) 1
1,net - bl b2 l/el + l/ez _ l
ql,net = (ebl - eb2) l/e
1
+ l/ez _ 1
For metal plates, we can improve Eq. (11.22) by the approximation ofEq.(11.14);
Similarly, we write the radiant flux from surface ~ to surface 3 as
we evaluate the emissivity of the cooler surface at T'I' = ft(i;.. For example, if
surface 2 is hotter than 1, we write Eq. (11.22) as 1
q -(e -e ) - - - - -
2,net - b2 b3 l/e 3 + l/ez _ l
1
ql,net = (ebl - eb2) l/e (T*) + l/ez _ l (11.23) Since.q 2,net = q 1,net (steady state), we eliminate eb 2 from these equations:
1
(eb1 - eb3)
We accomplish the evaluation of radiant heat transfer between nongray surfaces ql,net = (1/e 1 + l/e 2 - 1) + (l/e 2 + l/e 3 - 1).
with the most precision only by considering the best approximation for their
absorptivities, and, also the variation in properties with direction. Fortunately, . The ratio of radiant flux with the shield to that without the shield is
many industrial materials, especially refractories used in furnaces, behave as gray q 1,net (with shield)
diffuse radiators so that these details may be omitted from analysis. We shall
q 1,net (no shield) (l/e 1 + 1/e 2 - 1) + (1/e 2 + 1/e 3 - 1)
consider nonblack surfaces as gray in this chapter. More details regarding nongray
bodies and directional properties can be found in more advanced treatments of With i; 1 = i; 3 = 0.8, and i; 2 = 0.2, this ratio equals 0.285. Insertion of more shields
radiation heat transfer. would lower the ratio even more considerably.
l
~--!
---------new-1acrors~--~-----
.
dE12 (hemisphere) = 2nl
f
rt/2
2
cos 1 sin 1 r d</J 1
r2 2
, Btu/hr-ft ,
or
Surface 2--,..,....
dE12 (hemisphere)= nl. (11.27)
Note that
. dE12
Fig. 11.12 Exchange of radiation between elemental areas. dE1 2 (hemisphere) = - - = e 1 . (11.28)
dA 1
I
l_
- - - - - - - - - - - - . -.. ~----- ..
,,..,."'D------.::rr-:~-----
Since we have been discussing a black body, the result states that the emissive At thermal equilibrium, Qi.net = 0, and from Eqs. (11.32) and (11.33) and the fact
power of a black body eb is the product of n and the intensity of the emitted radia- that the emissive power of a black body depends only upon temperature, we see
tion: that A 2F21 = A 1F12 . In general, we can say that
(11.29) A;Fii = AjFJi (11.35)
Equation (11.35), though simple, is very useful, and the reader should keep it in
Thus, by combining Eqs. (11.26) and (11.29), we return to the major argument, mind.
and express the general equation for the transfer of radiant energy from one black
Values of F 12 have been calculated for various geometric arrangements of two
body to another as
surfaces. Figures 11.14-11.16 provide such view factor charts. More extensive
_ Jd
E12 - Ei2 -
_ eb
n
J Jcos </J 1
r
zcos </J dAz dAi.
2
(11.30) listings can be found in texts on radiation heat transfer. Although we have restricted
the discussion leading up to the view factor charts to black surfaces, it is apparent
that for nonblack surfaces, the emissivities of which are independent of emission
For convenience, the view factor F 12 , which is dimensionless, is introduced
~
here; it is defined by the equation "": 0.5
AiF12 -- -1
n
J Jcos <Pi cos <P2 dAz d
r
2 Ai. (11.31)
....
0
t> 0.4
...::!
L=O.I
Ai Ai
This integral contains nothing but geometrical aspects of radiant exchange, and
has fortunately been evaluated for many geometries, and one does not have to
perform the integration oneself. Now we can write a simple equation to express
the energy emitted by a black surface 1 and intercepted by surface 2, using the
definition of F 12 ,
l/.~
.~
>
0.3
0.2
O.I
0.4~-
0.6 - - - - - -
1.0
2__---::---
4-::::::::==== =
6
.
10 20--=
-'
1.0 2 345 IO 20 oo
(11.32) : 1.0 (c) </J = 90 lf
: 0.9 "": 0.22
In exactly the same manner, we could have developed an analogous expression for ~ 0.8 0 0.20
...::!
the energy originating at 2 and arriving at 1. " 0.7 t> 0.18 0.2
;; 0.6 ...::! O.I6
~ O.I4
......---- -
E21 = eb2A2F2i (11.33) 0.5
0.4 > 0.12 0.4
0.6 - - - - - -
Then, the net exchange of energy from 1 to 2 is Aiqt,net or Q 1 ,net: 0.3
0.10
0.2 0.08
(11.34) 0.06
Qi,net = E12 - E21 O.I 0.04
IO ... 0.2 0.3 0.5 1.0 2 3 4 5 JO 20 0.02
M
l( 0.7
1.0 - (a) <P = 30
0 ~0~.1~0~.2~0~.3~0.5~1.Eo=::::::2:::::3=4=5~~Io=:::2:::::::0
....
B 0.5 (d) <P = 120 lf
u
...::! 0.4 : 0.7 : 0.06
L = O.I
>""' 0.2
0.3
""s o.6
u
0.2 ""s o.o5
u
0.1 ..e o.5 0.4~- ...::!
0.07
0.05 "
;; 0.4 0.6~-
>"'
" 0.04
1.0 2 3 45 10 20
-'
O.I 6 IO-
2 0-
---
om 46 I 0 - - - =
y
I b
x = b /c and y = a/c
0
0.1 0.2 0.3 0.5 1.0 2 3 4 5 IO 20 oo
20--
0.2 0.3 0.5 1.0 2 345 IO 20 ~
(b) <P = 60 lf (e) </J = 150 l'I
Rwliaaw "?""'if",
Fig. 11.14 View factors for identical, parallel, directly-opposed rectangles. (From T. J. Love,
ibid.)
Fig. 11.15 View factors for two rectangles with a common edge. (From T. J. Love, Radiative
Heat Transfer, ibid.)
.-
--~---.,...,. .. - - - - - - - - - - - - - - - - - - - - - - - - - - ------------------~------~----
.;: 1.0 Similarly, the energy emitted by surface 2 and intercepted by surface 1 is
M
0
ti
~
E21 = eb2A2F21
~
The net exchange of energy from surface 1 to surface 2 is
>"' 0.7
Ql,net = E12 - 21
0.5 and, since A 2F21 = A 1 F 12 , we cari finally write the net exchange of energy from
E = r2 /d and D = d/r 1 0.4 surface 1 to surface 2 as
0.3 4 4
Qi.net= A1F12(eb1 - eb2) = A1F12a[T1 - T2 J.
Figure 11.15(c) is used to evaluate F 12 with
0.1
E=0.2-
0.2 0.3 0.5 LO 2 3 4 5 10 N=~= 3
12
L=-=2 and 4> = 90.
D
6 ' 6 '
Fig. 11.16 View factors for parallel, directly-opposed disks. (From T. J. Love, Radiative Heat The view factor F 12 is 0.165. Recalling that a= 0.171 x 10-s Btu/hr-ft 2 R 4, we 0
angles, F 12 calculated by the method discussed above will again represent the
fractional radiation from A 1 intercepted by A 2 . Thus such a calculation is exact
for black surfaces, and very useful for most nonmetallic, oxidized or rough metal
Ql,net = (72)(0.165)(0.171)[ (\~~or - (~~~ rJ
surfaces which exhibit nearly isotropic surface properties.
= 76,000 Btu/hr.
Since the view factor is the fraction of radiation emitted by a surface and
intercepted by another surface, it is obvious that all the radiation from any surface 11.7 ELECTRIC CIRCUIT ANALOGY FOR RADIATION PROBLEMS
must be ultimately intercepted when there are many surfaces involved. For
example, if surface 1 is seen by surfaces 2, 3, 4 ... n, then Consider an all-black surface enclosure (Fig. 11.17a) in which
Example 11.2 Calculate the net heat flow (Btu/hr) by radiation to the furnace wall The net heat flow (Btu/hr) from surface 1 is
at 500F from the furnace floor at 1000F. Both surfaces can be considered to be Ql,net = (E12 - E21) + (E13 - 31) + (E14 - E41)
black radiators.
= AiF12(eb1 - ed + AiF13(eb1 - eb3) + AiF14(eb1 - eb4). (11.38)
I
ebz A z F13 eb3
Side wall (500 F ),
surface 2
3
'II HJ--42--~/ ,~______,,.....,3 HIi .
00
Furnace floor, 2
surface I (I 000 F)
4
12'
1 ...
Solution. The energy emitted by the black surface 1 and intercepted by 2 is (a) (b)
E12 = eb1A1F12 Fig. 11.17 (a) A black enclosure. (b) Its electrical analog.
r-urnace-enc1o~es--------::J8'1
We can also develop Eq. (11.38) by analyzing the situation with an electric analog. Solution. The system and the appropriate analog are shown below.
We think of each surface as a node in the electric analog (Fig. 11.17b) and represent
r,~
the potential at each node by eb;, the current by the heat flow, and the resistance
between two potential nodes i andj by 1/A;Fii. We draw the circuit, and connect
each node by resistors to all the other nodes. We can then easily recognize the net
heat flow from any surface. For example, we obtain the net heat flow from surface 4
by simply summing the analog currents through the three resistors connected to
node 4:
Q4 net= AiF14(eb4 - eb1) + AzF24(eb4 - eb2) + A4F43{eb4 - eb3). ~
I - '1
t1
I
A, F 12
'--v---'
k 2
~
eb, J,
'ft may be desirable to express the net heat flow only in terms of A4 . In this case, J2 eb2
e Then by using Ohm's law, we can immediately write an expression for the net heat
qnet = - - - (eb - J). flow from 1 to 2 :
1 -e
Therefore, for the net heat flow from a surface of area A, we can write Ql,net = (ebl - eb2) 1 1 1
- e1 - e2
e --+--+--
Qnet = --A(eb - J). (11.42) A1e1 A 2 e2 A 1 F 12
1- e
In this case, A 1 = A 2 , and F 12 = 1. Therefore
Thus, we can represent a gray surface with a potential of a black surface
reduced by a resistance (1 - e)/ As, as in Fig. 11.18. The use of an electric analog is _ Ql,net
q 1,net -~
particularly usefuf when dealing with gray surfaces. We illustrate this in Example 1
11.3. 1
- (e e )------
I -t - . b1 - b2 l/s1 + l/e2 - 1
Ae
This expression is the same as Eq. (11.22). After having treated a few situations by
J
using electric analogs, the reader will become quite apt at developing expressions
Fig. 11.18 Electric analog of a gray surface.
involving radiation exchange between gray and/or black surfaces.
Fig. 11.19 (a) Furnace enclosure. (b) and (c) Its electric analogs. (In this case, conductances In furnaces, electric resistors very often provide the source of heat, and can be
rather than resistances are indicated.) mounted on a back-insulated refractory wall in a parallel array, as shown in
electric resistors to heat a metal part placed in a furnace. In normal furnaces, the Fig. 11.20. It is convenient to replace such a system by an equivalent continuous
radiation incident on the refractory walls is so large compared to the heat conduc- gray plane, having effective emissivity Gp and temperature equal to that of the
tion through the walls that we may approximate the conduction as zero when resistors. The gray plane is really imaginary and we show it by the dotted line in
calculating radiant fluxes. Such walls are examples of no-net-flux surfaces. This , I Fig. 11.20. To calculate Gp, we make use of Fig. 11.21 to determine FPE, this is
assumption greatly simplifies the problem of transferring radiant heat from surfaces equivalent to F12 in Eq. (11.43). Then, by rearranging Eq. (11.44) with Gp = 1, we
to sinks. can determine ffPE from the expression
As an example of this problem, consider the transfer from face 1 to the opposite
face 2 with intermediate refractory walls (Fig. 11.19a). All four walls are assumed to (11.46)
be gray, and the electric analog is given in Fig. 11.19(b). First, simplify the circuit
by combining all the resistances between 11 and 12 , remembering that in parallel
circuits conductances are directly additive, and in series circuits the same applies
to resistances. The equivalent conductance between 1 1 and 12 is where C and Dare defined in Fig. 11.21. If no reradiating surface (no back-insulated
- 1 refractory wall) exists, then ffPE = FPE in this expression. Finally, we can consider
A.1F12 = A1F12 + l/A1F1R + l/A2F2R' (11.43) the elements equivalent to the plane Ap with an emissivity Gp = :FPE at the operating
temperature of the elements.
and the circuit reduces to Fig. 1 l.19(c). The symbol F12 is often used when the
several view factors in a system are all combined, as illustrated here. Finally, the Example 11.4 A furnace has a floor 3 ft x 6 ft, and located at the top, 5 ft away,
equivalent resistance for the entire circuit is are heating elements, 2 in. in diameter, 3 ft long on 4-in. centers backed by a w((ll-
1 1 1 - GJ 1 1 1 - G2 insulated refractory roof. The heating elements are Nichrome IV (G = 0.74), which
- - = ---+---+--- (11.44) operate at a surface temperature of 2000F. The side walls are insulated refractory
A1ffei2 A1 GJ A1F12 A2 G2
i
__.L
.. ...-----
r
--------~---~aurn~.- ....,--..-.,.. -,....E:r.._,.----------- ------ ---------------------...---~
Second row
First row
r--
.
N o-ne t-flux walls
:~ x
~~
0.6
0.5
0.4
""
(no-net-flux) surfaces. Find the rate at which heat can be transferred to a cold sheet
of aluminum (3 ft x 6 ft, s = 0.15) placed on the floor of the preheated furnace. ::: 1.00
I~ 0.90
0
Solution. To demonstrate only the aspects of radiant heating, we shall ignore ] 0.80
convective heating on the upper surface ofthe aluminum sheet and any heating that
~ 0.70
might occur by conduction across minute contact points between the sheet and the
-5"' 0.60
furnace floor. "'
~ 0.50
To calculate the rate at which heat is transferred by radiation to the upper
surface of the sheet, we divide the system into four zones: (1) an equivalent gray ~ 0.40
0.30
plane to represent the elements, (2) the side walls of dimensions 5 ft x 6 ft, (3) the
side walls of dimensions 5 ft x 3 ft, and (4) the upper surface of the aluminum - Scale changes here
sheet. The analog circuit (showing conductances) i~ 0.10
I
o.....__._'-'-...._._._.L....J.......L__.._~....__.___.._~....__.___..__,
0 0.2 0.4 0.6 0.8 1.0 2 3 4 6
Ratio, diameter or least width D
thickness of wall X
(b)
Fig. 11.22 Total exchange factors between parallel source and sink connected by no-net-flux
walls. (a) Various geometries included and (b) values of the total exchange factor. (From
H. C. Hottel and A F. Sarofim, ibid.)
I
In this particular problem, to solve for A 1 F14 , the equivalent conductance for sheet interface is via radiation across a gap of two infinitely long parallel plates.
the entire circuit between J 1 (source) and J 4 (sink), would be a messy task if done by Using Eq. (11.22), and assuming that the emissivity of the refractory floor is 0.8, we
algebraic means. A 4 x 4 matrix could be set up and solved on a computer rather obtain
easily, or one could set up an actual analog circuit. For our purposes, we refer the
reader to Fig. 11.22, from which F14 can be found directly. In Fig. 11.22, the total ' (6 x
"---- -- 3)(0.111{(21~600r - (~~~rJ
exchange factor is presented for several geometries of source-sink combinations
QR,net = ---1/-0-.8-+-1-/0-.1-5---1---
connected by perfect reradiating walls. (Naturally, we have selected a geometry
represented on the figure!) By using Fig. 11.22, we can represent the circuit very = 165,000 Btu/hr.
simply as
Therefore
Qtotal = 141,000 + 165,000 = 306,000 Btu/hr.
The heat flow rates calculated above are the initial values, and apply only
when the cold sheet has been inserted in the oven. We deal with the more general
problem of describing the transient heating-up of the sheet in Section 11.12.
Using Fig. 11.22, we find, for the geometry of a 2:1 rectangular opening with
D = 3 ft and X = 5 ft, that 11.9 RADIATION COMBINED WITH CONVECTION
F14 = 0.48. In the preceding sections of this chapter we considered energy transfer by radiation
Next we proceed to determine 8 1 . From Fig. 11.21 (curve 5) with C/D = 2, we as an isolated phenomenon. Indeed, in many high-temperature applications,
obtain FPE = 0.88. Then, from Eq. ( 11.46), problems should be tackled in this manner because, at high temperatures, radiation
can completely dominate since radiant heat flow depends on the fourth power of
= = ) = 0.741. the absolute temperature. In many practical situations, however, we cannot neglect
8 1(8p) :ff'PE
1 2( 1 convective heat transfer, and it is necessary to consider both modes of energy
-+- --1
0.88 n 0.74 transport.
Now using Eq. (11.44) with A 1 = A 4 , we get When we include radiation in a calculation of a situation which also involves
convection, we realize that the total heat flow is the sum of the convective heat
1 1 - 81 1 1 - 84 flow an~ the radiant heat flow. Thus, we conveniently use the total heat-transfer
-=--+-+--
ffei4 81 F14 84 coefficient ht, which is composed of the heat-transfer coefficient h in the usual sense,
and the radiant heat-transfer coefficient h,:
1-0.741 1 1-0.15
0.741 + 0.48 + 0.15 ht= h + h,. (11.47)
Therefore
(11.48)
4 4
Ql,net = A1ffei4rr(T1 - T4 )
I
.- .- -.-.-- . --------------------~...--------r~,...-------~---------~. ,~ -.
Lt 100.0 Solution. The temperature of the thermocouple is less than that of the gas because
90.0 I the thermocouple radiates to the wall. We can write an energy balance for steady-
80.0
state conditions in which the radiant heat flow from the thermocouple to the wall
I
! equals the convective heat flow from the gas to . the couple. Assuming that the
thermocouple can be approximated as a gray surface, we write
50.0
hA 1 (T1 - T1 ) = A 1 F 12 e 1 a(T14 - T24 ),
where A 1 is the thermocouple surface area; T1 , T1 , al)d T2 are the temperatures of
30.0 the fluid, thermocouple, and duct walls, respectively. Since the thermocouple is
completely surrounded by the duct walls, F 12 = 1. The gas temperature is then
found by
. e1 a 4 4
T1 - T1 = h(T 1 - T2 )
4
= (0. 7)(0.171) [ ( 1460)
(20) 100
7.0 = 222R or F,
6.0
and
T1 = 1222F.
4.0
Fr= o.m[(~)4-L~o)J Errors such as those indicated by the above problem have induced engineers to
T 1 -T2
T = degrees Rankine pursue thermocouple designs which incorporate radiation shields and/or means of
increasing h between the fluid and the thermocouple. Some devices are illustrated
in Fig. 11.24.
2.0
Example 11.5 A thermocouple with an emissivity of0.7 measures the temperature The methods given in the preceding sections are applicable only to systems involv-
of a gas flowing in a long duct whose internal wall surfaces are at 500F. The ing gases which are transparent to radiation. Gases with simple symmetrical
temperature indicated by the thermocouple is 1000F, and the convective heat- molecules, such as H 2 , He, 0 2 , and Nz, are essentially transparent to radiation, but
transfer coefficient between the gas and the surface of the thermocouple is heteropolar gases (CO, C0 2 , H 2 0, S0 2 , NH 3 , HCI), which interact with radiation
20 Btu/hr-ft 2 F. Determine the true temperature of the gas. sufficiently, must be made part of the system for calculations.
1
Solids emit radiation at all wavelengths, whereas gases emit and absorb radia-
6on only between the narrow regions of wavelengths called bands. Analysis of
radiation participation should be made for each band, and the total effect can be
obtained by summation. However, it is more convenient to solve the problems in
terms of the total radiation and emission characteristics of the gases.
As a first approximation, the total number of radiating gas molecules present
in a space determines the emissivity and absorptivity of the gas. This approxima-
tion is best at lower pressures with corresponding large molecular mean free paths
permitting each molecule to radiate as though it is alone, with a minimal shielding
effect by the other molecules present in space. Gas radiation differs from the
i radiation of solids whose emission and absorption are surface phenomena, because
when calculating emission and absorption for a gas layer at a particular tempera-
ture, its thickness and pressure must be accounted for. In fact, the emissivity
and absorptivity of a gas is a function of the active gas species, temperature,
thickness and shape, total pressure, and the partial pressure of the active gas. To
handle all these complexities, a series of charts has been developed based on
experimental data for evaluating the emissivities and absorptivities of various gases.
Consider a hemisphere of radius L, containing a heteropolar gas of a given
partial pressure; the problem is the radiant heat exchange between the gas at
temperature ~ and a unit element at the center of the base of the hemisphere. The ' Absolute temperature, 0 R
emission of the gas to the surface is t,gr;~4 per unit surface area, where eg denotes Fig. 11.25 Reduced emissivity of carhop dioxide. (Figures 11.25-11.27, 11.29, and 11.30 are
gas emissivity. Specifically, in Figs. 11.25-11.27 the reduced emissivities e~ for adapted from H. C. Hottel and A. F. Sarofim, Chapter 6, ibid.)
C0 2 , CO, and H 2 0, are presented as a function of the absolute temperature and
the product of Land partial pressure. These values are called reduced emissivities
because they are plotted for a reduced pressure; that is, all data are at near-zero >. 0.15
partial pressure and at a total pressure of 1 atm. The graphs apply strictly to the -~
r:i 0.12
hemispherical body of gas defined above; we shall discuss briefly how to treat other ~ 0.10
shapes. c3 0.08
The effect of total pressure on eg for C0 2 and H 2 0 is given in Figs. 11.28 and 0.07
0.06
11.29, where the ratio of the actual emissivity, at the state being considered, to the
reduced emissivity is presented as Cc and Cw for C0 2 and H 2 0, respectively.
When both H 2 0 and C0 2 are present in the gas, we must make another correction
for the mixture composition using Fig. 11.30. We subtract the value of ~t: in this
chart from the sum of emissivities for each component. This correction is due to
the fact th.at both gases emit radiation in overlapping wavelength bands.
For shapes other than hemispheres, we take Las the effective beam length, and
calculate it for use with Figs. 11.25-11.30. Table 11.2 lists L for shapes other than
hemispheres. For approximate calculations of shapes which are not listed, we can
take Las 3.4 x volume/surface area. For a presentation and discussion of radia-
tion data for many other gas species, the reader should consult Chapter 6 of the
text by Hottel and Sarofim. 4
Temperature, 0 R
4
H. C. Hottel and A. F. Sarofim, ibid. Fig. 11.26 Reduced emissivity of carbon monoxide.
..,-------~:::r~--.n.-uv1un,_,.1-.1-.;:;-a-.--.-.-n~ .."'- - -
- - - -- - -.
----r--~
------------------------~-~..,.~=-------
>.
.->;::
>
-~
0.7
0.6 ~
a 0.5
"
00
0"'
0.15
Absolute temperatu;e, 0 R
Fig. 11.28 Correction factor for C0 2 emissivity. (From F. Kreith, Principles of Heat Transfer,
ibid.) Fig. 11.30 Correction for band overlap when C0 2 and H 2 0 coexist.
I
l
Table 11.2 Beam Jen.gths for gas radiation. (From H. C. Hottel, "Radiation," Chapter IV of Then
Heat Transmission, by W. H. McAdams, McGraw-Hill, 1954)
PcL = (0.40)(2)(1.8) = 1.44 atm ft,
Factor by which X is
multiplied to obtain
PwL = (0.10)(2)(1.8) = 0.36 atm ft,
mean beam length, L !(P + Pw) = !(2 + 0.20) = 1.10 atm,
Shape Characterizing
dimension, X For average and
When values of 0.20 = 0.20.
PgL = 0 PgL 0.80 + 0.20
2/3 0.60 From Figs. 11.25 and 11.27-11.29, s~ = 0.17, s~ = 0.12, Cc= 1.10, Cw= 1.55,
Sphere Diameter
Diameter 1 0.90 and .1s = 0.032.
Infinite cylinder
Semi-infinite cylinder, radiating to center
Diameter 0.90 Therefore
of base
Right-circular cylinder, height = diameter,
radiating to center of base Diameter 0.77
Same, radiating to whole surface Diameter- 2/3 0.60 = (1.55)(0.12) + (1.10)(0.17) - 0.032 = 0.34.
Infinite cylinder of half-circular cross sec-
tion. Radiating to spot on middle of fiat
Radius 1.26 11.11 ENCLOSURES FILLED WITH RADIATING GASES
side
Rectangular parallelepipeds By evaluating the emissivity and absorptivity ofa gas by the method outlined in
1 :1 :1 (cube) Edge .2/3 Section 11.10,.. we can evaluate the interchange of radiant energy between a gas
1 : 1 : 4, radiating to 1 x 4 face } 0.90
and the surface of its enclosure. First, consider a black bounding surface. The gas
radiating to 1 x 1 face Shortest edge 0.86
emits enei;gy which is entirely absorbed by the black surface, but, on the other hand,
radiating to all faces 0.89
1.18
the gas absorbs only a fraction of the radiation from the surface. Thus we write the
1 : 2: 6, radiating to 2 x 6 face } flux of energy from the gas g to the surface 1 as
radiating to 1 x 6 face 1.24
Shortest edge
radiating to 1 x 2 face 1.18
radiating to all faces 1.20
1 : oo : oo (infinite parallel planes) Distance between 2
planes 4 4
= a(sg 1'i: - ocg 1T1 ). (11.50)
Space outside
Space outside infinite bank of tubes with When the bounding surface is gray, it reflects part of the energy emitted by the
centers on equilateral trian,gles; tube gas which is partially absorbed in successive passes back through the gas. In this
diameter = clearance Clearance 3.4 2.8 case, we give the flux of energy by the right side of Eq. (11.50) multiplied by a
Same as preceding, except tube diameter = fai;tor between s 1 and 1. For surfaces with s 1 of 0. 7 or greater, it is sufficient to
one-half clearance Clearance 4.45 3.8
use the following equation:
Same, except tube centers on squares;
diameter = clearance Clearance 4.1 3.5 S1 +
qg .... 1 = a ( --2-1)
(sg1'i:4 - ocg1 T14 ), (11.51)
Example 11.6 Determine the emissivity of a gas containing 50% N1: 40% C:02,
where the factor is simply (s 1 + 1)/2.
and 10 % H 20 at a total pressure of 2 atm with 1'i: = 2000R. The gas is con tamed
In order to incorporate participating gases into a system containing surfaces,
in a very long cylindrical space having a 2-ft diameter.
we consider the gas to be a floating node of potential ebg and of conductance A;sg;
Solution. From Table 11.2: between the gas node and the surface i. Due to the presence of the gas which
L = (0.90)(2) = 1.8 ft. absorbs some of the radiation between surface nodes, the conductances between
-------nans1enrcmmuctron w1thra1mnron-arme surface 3
the surfaces themselves are multiplied by the gas transmissivity; for example, the Now check the assumption that ~ = 1000F.
conductance between surfaces i and j is A;Fijrigi The following example demon-
4
eb = aT{ = (0.1713)(16.6) = 13,000; similarly, = 4300 Btu/hr/ft 2.
strates the use of such an electric analog for a system with a participating gas. 1 eb 2
Example 11.7 Two large parallel plates (with emissivity of0.6) at temperatures of Then
1200F and 800F transfer radiation across a 3-in. gap filled with C0 2 at 1-atm
pressure. Calculate the rate of heat transfer from the hot plate to the cold plate. 11 = eb 1 - / 1 -~ = 10,590 Btu/hr ft 2,
s 1 (1 - si)
Solution. We show the system and the equivalent network below.
12 = eb2 + t-
s 21 1 -
2
B2
) = 6710 Btu/hr ft2.
Since sg 1 = Bgz, the potential ebg must be the average of 11 and 12 , so that
10,590 + 6710
ebg = = 8650 Btu/hr ft 2,
2
and
I
4
The problem could be solved again with this as the gas temperature; in this case,
however, 1030F is sufficiently close to the assumed 1000F, so that no significant
improvement in the result would be gained.
Figure-Example 11.7: (a) Radiation between parallel plates with a participating gas, and
(b) the electric analog. It is instructive to extend the above example, by comparing it to the case of no
participating gas (see Example 11.3). In this case
In order to evaluate the emissivities and transmissivities of the gas, we must know
its temperature. Assume the gas temperature is 1000F (this can be checked later). 1 1 1 1 1
-=-+--1=-+--1
From Table 11.1, L = (2)(3/12) = 0.5 ft. Then, PcL = (1)(0.5) = 0.5 ft atm, and ffi2 81 82 0.6 0.6
from Fig. 11.25, Cg = 0.12; the correction for nongrayness is small in this particular = 2.33.
problem, so that Bg 1 = Bg 2 = IJ( 1g2 = 0.12, and r 1g2 = 0.88. The network can be
reduced to a single equivalent resistance for current (heat flow) between 1 and 2. This is hardly different from the factor (2.39) determined in Example 11. 7. Therefore,
Here A 1 = A 2 = A and F 12 = 1, so that even though we have selected an example involving an atmosphere of 100 % C0 2,
we could still consider the atmosphere to be transparent for many calculations.
1 1 - 81 1- 82 Only at partial pressures greater than about 4 atm of the participating species
-=--+ +--
ffi2 81 1 82 and/or with very large enclosure dimensions, would the effect of the gas be more
F12'1g2 + - evident.
1/sg 1 + 1/sg 2
~The likely application of these principles to metallurgical processes, which
0.4 0.4 ordinarily involve combustion gases at 1 atm total pressure, requires only small
= 0.6 + 1 + 0.6 corrections to heat fluxes. For example, a typical combustion gas at 3000R could
(l)(0. 88 ) + 1/0.12 + 1/0.12 contain 12 %C0 2 and 20 % H 20. From Figs. 11.25 and 11.27, the reduced emissivi-
ties for a rather large L (e.g., 3 ft) are only 0.08 and 0.11, respectively.
= 2.39.
Then,
11.12 TRANSIENT CONDUCTION WITH RADIATION AT THE SURFACE
qi-2 -- Ob
~11<J
(T4
1 - T24) -- (0.1713)(16.64 - 12.24) -- 3620 Btu /h r ft 2. In Chapter 9, we discussed some solutions to transient problems involving solids
2.3 9 with heat applied to, or withdrawn from, the surface via convective h. In several
- - - - - - , " K I > ; - - -. . . .
-----------------------------,-~~--------
operations, notably heat treating processes, stock is heated via radiation. The f..,o 10
I 111 I
I
v
condition of heat input at the receiver (stock) surface from a heat source at constant Plate 1/0, = T 0 /T, j I
""'"
8 1 - - - k ->oo Tl
temperature T, is
J-s 0.10I ~
ff <J[T4 - T4(0 t)] = - k oT(O, t). (11.52) 6 /
sl s ' OX I / 0.15
4
-/ I/ v 0.20"
Here all temperatures are absolute, and we determine ffe. 1 according to the methods I/
~I~ ~ t:: 1- -0.30- o.so~ I/
outlined above, depending on the spatial relationship between the source and
stock and their respective emissivities. We can write Eq. (11.53) in the dimensionless 2
~ ...-
' / ' vl..---"""c040
~ i..--' I~
~
at
M= ffe, 1 <JTi
k
l,
Fo =-
12 ~ 1.0 l
~ ~f:'
......... -.r- JI1.5
~"to..
#
h
e=-
T T,
es = To
II
"" 0.8
r--,-... ,.......
"'"' :---.-
....,
I'-~
~ "'"
- ~
['1' 2 1'--r-
To ['\. 3 -- 5~
"'"
~
Plate
0.(
,~
L-- k ~ 00
T0 is the initial temperature of the stock, and l is its characteristic dimension. "f::: "~ 1"':5 4 I"
7 6 ,1-.....1' ~ ['-..._ 1'--r-
Consider specifically the temperature response of a sheet, 0 ::; x ::; I, thin 0.4
"o, 8 ,, .....
Ii ~ I'-
-.........
enough for Newtonian conditions (no internal gradients) to apply, which is heated I 1'-r-
1/8,= T0 /T, f::: ~~~ ~~ ~ I:::
at x = 0 by a radiation heat source at constant temperature. The back face 0.2
(x = l) loses a negligible amount of heat so that [oT(l, t)/ox] = 0. We give the
solution to this problem in Fig. 11.3 l(a) and (b) for heating and cooling, respectively.
0.0 I Ill I
0 "'
0 0 "'
0 0 "'
0 0 ci"' c:i "'
ci
For other closely related situations we refer the reader to Schneider. 5 6 ci
0 0 0 0 0 0
0
0
ci
0
~
~
0
6 c:i ci
MO;Fo = (a.i'",,T;'/pCP/)I
0
Example 11.8 (from Schneider 5 ) We heat treat a continuous sheet of steel (0.041 in. (b) Cooling, 0, = T,/T0 <1
thick) by passing it through a radiant furnace designed to provide a heat-up and
cool-down cycle. The furnace consists of a 10 ft long ceiling heat-source section Fig. 11.31 Temperature response of a plate, 0 ~ x ~ I, with no internal thermal gradients and
(ffe. 1 = 0.72) at 2200F followed by a ceiling heat-sink section (ffe, 1 = 0.86) at an insulated back face, at x = I, after sudden exposure to (a) a radiation heat source (heating)
- 200F. If the sheet enters the furnace at 70F, must be heated to 2000F, and or (b) a radiation heat sink (cooling). (From P. J. Schneider, Temperature Response Charts,
then cooled to 500F before leaving the oven, calculate (a) the required feed Wiley, New York, 1963.)
3
velocity and (b) the length of cool-down section. Data: density of steel = 486 lq/ft ;
heat capacity = 0.11 Btu/lb-F. // a
Solution
Heating Cooling
a) For the heating section, we apply Fig. 11.3 l(a) with
(/SI
From Fig. 11.3 l(a) b) Calculate the rate of heat transfer (by radiation and convection) from a vertical surface
5 ft high and the percentage by radiation for the following cases:
M03Fo=fl'.1aTrfl_(T.)3 k t
s k T0 . pCP /2 Surface (t: = 0.8) at: Air at:
I
rr---------------------- --------------
a) Calculate the heat loss by radiation (Btu/hr) through the open hole. not only supplies power to melt the metal but also levitates it. But with this set-up, the strength
b) Calculate the heat loss if a sheet of nickel 6 in. x 6 in. is fitted across the hole on the outside offield necessary to keep the sample levitated is sometimes such that the metal gets overheated.
surface. A means of preventing this is to pass a cooling gas past the sample.
Develop an expression that relates the steady-state temperature of the metal to the heating
Data: emissivity of furnace refractories = 0.9, and emissivity of nickel sheet = 0.4. power supplied by the field (QP, Btu/hr), the gas temperature T0, the gas velocity V, and any
other parameters which you think are essential. Metal temperatures of interest are 2000-
[Note. The furnace refractories should be approximated as perfect heat insulators.]
30000F, and the convection h_eat-transfer coefficient can be expressed as
11.5 Cast iron is continually tapped from the bottom of a cupola into an open refractory
he= KV 0 7 ,
channel. The metal enters at 2800F and runs down the channel at a rate of 4000 lb/hr. The
dimensions of the channel are shown below. in which K is a known constant.
Neglect the heat loss by conduction through the refractory and estimate the metal discharge
11.7 A steel sheet tin. thick and having the shape of a square 5 ft x 5 ft comes out of a heat-
. ~' temperature.
treating furnace at 1500F. During heat treatment its surface was oxidized so that its emissivity
is 0.8.
6" a) Calculate its initial cooling rate if it is suspended freely by a wire in a room at 80F. Neglect
~1 all heat transfer with convection, i.e., deal only with radiation heat transfer.
b) Calculate its initial cooling rate if it is supported vertically on a horizontal surface (also at
Cupola
80F) which has an emissivity of 0.2. Again deal only with radiation heat-transfer.
Data (all units in Btu, hr, ft, F, etc.):
11.8 Brass sheet, / 6 in. thick, is to be tempered in a continuous tempering furnace. The first
portion of the furnace is filled with gases at 1200F, moving against the strip direction at a speed
\\ (i
Data:
CPbrnss = 0.10 Btu/lb-F CPgas = 0.26 l'/gas = 0.1 lb/ft-hr
Pbrass = 574 lb/ft3 Pgas = 0.03
Cooling gas kbrass = 100 Btu/hr-ft F kgas = 0.05
r
12
Distribution
plates
This chapter deals with the transfer of heat under turbulent flow conditions
between a packed column of solid particles and a gas flowing through the packing. Gas in
Counter-current flow or static-bed situations are the only cases we shall study, since
they represent virtually all the metallurgical applications of this type of heat trans-
fer. Metallurgical processes utilizing the counter-current :('k)w of solids and gases
include, for example, iron and lead blast furnaces, shaft pelletizing and/or calcining
furnaces, iron cupolas, and copper blast furnaces. The equations developed
for these cases can be easily modified to des~ribe the thermal behavior of crossflow
processes, such as iron-ore sintering and pelltjtizing and zinc-ore roasting and
sintering. In most of these cases, the complexity of the equatfons makes analytical
Pellets out
solutions almost impossible, and most studies which have attempted to do so
(a) Chlorination reactor
have ultimately relied on numerical integration methods to provide final results. (b) Iron cupola
404
l'
________ ____ ____
"_"_" ""-""""""_" "-""" --"-----~
;reauy--smTe;-coumer-curr.enrnow ""7
(12.2)
where S = total surface area of particles per unit volume of bed, ft 2 /ft3,
h = heat-transfer coefficient, between the gas and solids, Btu/min-ft 2 F, dz
QR = heat released by reactions, Btu/min-ft 3 ,
w = void fraction,
k0 rr = effective thermal conductivity of bed,
p., pg = densities of solid and gas phases, respectively, lb/ft 3 ,
c., cg = heat capacity of solid and gas, respectively, B_tu/lb F,
V0 g = superficial velocity of the gas stream, ft/min,
V, = actual velocity of solids, ft/min. ri
Temperature
For convenience in the following sections, we define two new' terms.
a) Thermal capacities per unit cross-sectional area of bed: Fig. 12.2 Notation and definitions for the analysis of counter-flow heat transfer.
3
Gs = PscsCI - w), Btu/ft F,
Gg = pgcgw, Btu/ft 3 F. Figure 12.2 illustrates the conditions and provides a reference framework for
the following development.
b) Thermal flows per unit cross-sectional area of bed: First, let us develop a simple expression for the length of bed needed to effect
a specified exchange of heat. In Figure 12.2, consider a slice of bed of thickness dz.
W. = V,Gs, Btu/min-ft 2 F, Within a unit area of that slice, the hot stream (the gases in a solid-"heating opera-
wg = VogGg/w, Btu/min-ft 2 F. tion) loses an amount of heat:
dq = - W,. dT,., Btu/min ft2, (12.3) .
_I
creaoy"sraTe;-coumer;currenrnow-<109
From Eqs. (12.3) and (12.4) we see that Applying these equations, and the relationship
1
d(T,. - T;) = - dq(- - __!__). (12.6) T,. = T~ at z = L,
l
l,, ~ we develop an equation for the temperature of the hotter stream, at any distance
Substituting dq from Eq. (12.5), we get z from its entrance:
1
d(T,. - T;) = - hS(T,. - T;)(- - __!__) dz,
l,, ~ T.(z) = T 0
1 - exp - -
hS (
l,,
1- - z
[
l,,)
~
J
1
h h
- (T 0
h
- T 0)
w. [
c 1 - _h exp - -hS ( 1 - _h L w.) J . (12.10)
or d(T,. - T;) = -hs(-
( T,. - T;) l,,
- __!__)dz.
~
(12.7) f ~ l,, ~
Integrating Eq. (12.7) with Similarly, the local temperature of the colder medium is
If a specified entering temperature for the cold stream, T~, is available, along with The equations given below are useful for
rh~(
TZ and T~, then we can calculate T~ for a given bed length (z = L), heat-transfer a) the outlet temperature of the hot stream:
coefficient, and thermal flow condition. On the other hand, if we predetermine all
the temperatures of the cold and hot streams at the ends o( the bed, then we find
the necessary bed length, since T~ ~ho
= - (Th
0
- Tc )
0
~f 1- W,
1 - - exp - -
W,)
1- - L
Jl; (12.12)
~ l,, ~
b) the outlet temperature of the cold stream:
and
l --
(12.9) Tt = To - (To - To)
c h h c W.
~exp[
~
hSl,, W.
~)LJ
l . (12.13)
This is a useful expression for design purposes, particularly where the only 1 1- - l,, { 1 -
specifications are on the final and initial temperatures, and the only design para-
Equations (12.IOH12.13) appear to be indefinite when l,, = ~, which is a
meter is the length of shaft desired.
condition often met in practice. We overcome this difficulty by applying L'Hopitals
On the other hand, there are situations in which we need to know the tempera-
rule, differentiating both numerator and denominator as l,,/~ --+ 1. The results,
ture profiles of the gas and the solid. This is often the case when it is necessary to
& for the special case where l,,/~ = 1, are then
know how long the solids are above some specified temperature. In order to
obtain the profiles, we use Eq. (12.8), and write it as
0
T,.(z) = Th - (Th - Tc ) 0 0( z W. ) , (12.lOa)
which gives the local temperature difference at any level z. We know, however, 1+ -
hSz)
from a heat balance between the entrance and any point in the bed, that T;(z) = Th 0
- (Th 0
- T~) ,:;~ ' (12.lla)
(1 + -
iv,,[n - Th)J = ~[T~ - T;(z)J. iv,,
,.,
~--,..-..---r=~------- ------------------~11eaoy;;sraYe;-coumerCDl"Tenrnaw~I"'I
.:::: 80
trg
"'"
u
<=
rg = ooF
"' T~ = 2700F
~
60 60
40 40
20 20.
0 500 1000 1500 2000 2500 0 500 1000 1500 2000 2500.
T~ T~
Temperature, F Temperature, F
Fig. 12.3 Effect of change in thermal flow ratio W,./W, on the region of heat transfer in a Fig. 12.4 Effect of change in the volumetric heat-transfer coefficient hS on the region of heat
counter-current packed bed. Volumetric heat-transfer coefficient hS is constant (2.5 Btu/min- transfer in a counter-current packed bed. Thermal flow ratio is constant at W,./W, = 0.75
ft3 F) in all cases. Note that the region of active heat transfer is pushed from one end of the for both cases. Note that in the case of the low coefficient, 1/i never reaches ~0 .
shaft to the other as the ratio changes.
Data:
L =
(558 _ 70)
2303 log (1500 I ( 1 1 )
- 1200) (0. 3)(20) 10.0 - 12.0
A nomogram has been developed, Fig. 12.5, which facilitates the determination
of hv.
For perfect spheres, the equation
12V. yo.3
= 4.75 ft. hv - Og
Di.35
(12.15)
p
12.3 HEAT-TRANSFER COEFFICIENTS IN PACKED BEDS agrees with Furnas' results and those of Saunders and Ford 3 .
Up to this point, we have presented the heat-transfer coefficient h, but we have not Example 12.2 Estimate the volumetric heat-transfer coefficient in a shaft furnace
mentioned how to obtain this value for a flow through packed beds. As might be in which hot gases (1200C) are used to heat harden iron oxide pellets, 1 in. in
expected, we do not obtain h in a packed bed from any of the correlations presented diameter. The gas velocity is 1.0 m/sec.
. t in Chapter 8; therefore new correlations are needed. Since
Solution. Entering the nomograph at 1.0 m/sec, going up to 1200C, across to
Nu = f(Re, Pr),
25 mm, and down, we find that hv = 12,500 kcal/m 3 -hr C. This becomes
is still true, we could look for an expression of the form
Nu = A Re" Prm. hv = 1.25 X 104 X 1.45 X 10- 3
Here we are dealing exclusively with gases as the fluid phase, and since the Prandtl = 18.5 Btu/ft 3 -min F.
numbers of gases are all about equal over a wide range oftelli.peratures, there is not
any strong dependence of Nu on Pr for gas-solid heat transfer. So we are left with 1800
finding the relationship between Nu and Re. However, this is much easier said 1600 l------ot.WI
than done, as most of the experimental work on this subject has been confined to 1 4 0 0 - - -...;
1200 1 --~,_,u
much lower temperatures than those of most metallurgical processes.
The highest temperatures-ll00C-were employed by Furnas 1 in his classic 1000
research in 1932. Recently, Kitaev and coworkers 2 have summarized the results
of the pertinent studies at elevated temperatures, and also some of the lower tem-
perature results. Their review, which is probably the best ever'made on the subject,
500
includes a complete reanalysis of Furnas' data. Since the effects of void volume
400
and temperature are so difficult to obtain, and as yet have not been firmly es-
300
tablished, the volumetric heat-transfer coefficient, hS or hv, is presented by them 200
in the form 100
A vg~9yo.3f(w) oo
hv = Do.15 ' (12.14)
p
where hv = hS = volumetric heat-transfer coefficient, k caljm 3 -hr C, 15,000 10,000 5,000 0 1.0 2.0 3.0
A = a coefficient dependent on bed material, Volumetric heat-transfer coefficient Gas velocity (0C)
h,,, kcal/m 3 -hr C in free cross section of bed, m/sec
T = temperature, C, (converted to STP)
DP = particle diameter, mm,
'Fig.12.5 Nomograph for the volumetric heat-transfer coefficient hS or hv, based on knowledge
v0 == superficial gas velocity, m/sec, of the superficial gas velocity V0 g, the gas temperature, and the particle diameter DP. (Adapted
f(w) = function of void fraction. from B. I. Kitaev et al., ibid.)
For natural, lump, materials, A = 160 andf(w) = 0.5, so that
8ovgi:9 yo.3 Multiply by to obtain
hv = Do.15 . ft/sec 0.3048 m/sec
p
Ill. 25.4 mm
1
C. C. Furnas, U.S. Bureau of Mines Bull., 261 (1932). kcaljm 3 -hr C 0.001045 Btu/ft 3 -min F
2
B. I. Kitaev, Y. G. Yaroshenko, and V. D. Suchkov, Heat Exchange in Shaft Furnaces,
Pergamon Press, New York, 1968. 3
0. H. Saunders and H. Ford, J. Iron & Steel Inst. I, 292 (1940).
12.4 STATIONARY BED, INFINITE HEAT TRANSFER
r
I
I Combining Eqs. (12.16) and (12.17), we arrive at
z-direction, perpendicular to the bed surface. Ifwe pick out a unit area of bed as it which is recognized as a quasi-linear partial differential equation of the first order,
moves into the gas contacting zone, we may analyze its unsteady-state behavior with the form
(ar.) (a r.)
as if it were stationary, at least until it passes out of the zone at the other end. In
Pat+Qa;=R. (12.19)
this situation (Fig. 12.6), JI, is zero.
where U 1 (Ts, t, z) = Clo and U 2 (Ts, t, z) = C 2 are the solutions of any two of the
following relationships:
dt dz dTs.
p Q R
and
dT, = 0 . dt = 0. (12.21)
Fig. 12.6 Schematic diagram of a cross-flow system, with the packed bed moving from left to
right, and the gases being drawn through it from top to bottom. Integration of Eqs. (12.20) and (12.21) results in
Let us consider first the situation where the assumed conditions, in addition
to those in Section 12.1, include:
z -(
Gg
~
+ G.
)t = U1,
(ar.)
with dry, heated nitrogen. The nitrogen enters at the rate of 1 lb/min for every
G. at = 0. (12.17) 3 lb ofFe 3 0 4 to be heated. The bed density is 150 lb/ft 3 , and the void fraction 0.52.
-r------
1
I= 0
t = l.14min
0 500
Temperature. F Temperature, F
Fig. 12.7 Propagation of a thermal front through a bed of solids. T,;0 is constant at all times. Fig. 12.8 Propagation of a thermal wave through a bed of solids. Tg0 decreases to Tt after
a) Illustrate the thermal profile of the solids in the bed at time zero, at the time 20 sec.
when all the Fe 30 4 is heated to 1800F, and when 50 %of the charge is heated,
assuming an infinite heat-transfer coefficient. Suppose that the temperature of the incoming gases, instead of being kept at
b) Calculate the time to heat all of the Fe 30 4 to 1800F. 1800F, has dropped to 800F after 20 sec. The profile at the same times would be
as shown in Fig. 12.8. Here the inlet gas temperature changes; after the change is
Solution. In Fig. 12.7 we see the answer to part (a). A vertical temperature front effected, the resultant profile propagates, unchanged, through the bed.
passes through tqe bed at a velocity [W8 /(G 8 + Gs)J. From the definition of the
example, we can calculate this velocity. 12.5 STATIONARY BED, INFINITE HEAT-TRANSFER COEFFICIENT, AND
HEAT OF REACTION
=0 2 ( Btu )( llbN 2 150lbFe 30 4 )
Wg .9 0 2 Let us now assume the same conditions as in Section 12.4, except that a reaction
lb-N 2 F 3 lb Fe 30 4 mm ft
liberating QR Btu/min ft 3 occurs in the gas phase within the bed. The energy
= 14.5 Btu/min-ft 2 F. balance equations, in terms of T,, are thus
c w = (0.017 lb 0.29 Btu 052) = 0.00256 Btu/ft 3 F,
G =
g Pg g ft 3 lb F lg aT.) + Gg(aT.)
(Jz at - QR = 0,
0.22 Btu 150 lb Fe 30 4 and
Gs = PsCs(l - w) = lb op ft3 bed
Gs aT.) = 0.
(at
= 33.0 Btu/ft 3 F.
Therefore, the velocity at which the front moves is Since we may combine the above two equations to yield
14.5
_ + ~o = 0.44 ft/min,
aT.) + (Gg w.+ Gs)(aT.)
(at QR
Jz = Gg + G,'
(12.23)
33 0
and the time, t1 , to heat the entire bed to 1800F is we can obtain a solution in the same manner"as shown in Section 12.4. In this case,
1 ft . = 2.28 min.
. U 2(T, t, z) = T, - (Gg+QR Gs )t,
I
0.44 ftmm .
\ - - - - P a c k e d bed - ----~
Solution. To start with, let us calculate the velocity of the front. This in tum re-
..... quires that we recalculate ~'using the heat capacity of air, 0.28 Btu/lb F, rather
.... than that of nitrogen. The difference is slight, and
w,,
,. =cg+c-; ..... Wg = 14.0 Btu/min-ft 2 F.
As before, Gs = 33.0 Btu/ft 3 F, and we will neglect Gg.
..... Therefore, the velocity at which the front moves is
12
..... -14.o = 0.424 rtI mm,
.
t,
..... and
33
Distance z. ft 1 .
.,_' '{1
tJ = 0.4
= 2.36 mm .
' Fig. 12.9 Temperature profiles in a stationary bed with gas at T;. entering a bed with an initial 24
hot zone, and heat generation rate QR. Note that tan a= QR/(Gg + G.). Now, in order to obtain "thermal front" temperatures, we need to evaluate
QR. The heat of oxidation of Fe 3 0 4 to Fe 2 0 3 is 210 Btu/lb Fe 3 0 4 .
).
and 150 lb Fe 3 0 4 210 Btu
( Gg ~
QR = ft 3 lb Fe 3 0 4 5 min
U 1 (T, t, z) = z - + Gs t,
= 6300 Btu/ft3-min.
so that the general solution is now Therefore
T =
s
( QR
Gg + Gs
)t + f [z -(Gs wg+ Gg )t] , (12.24)
Gg
QR
+ Gs
6300 Btu/ft 3 min
33 Btu/ft 3 F
Thus, any initial or impressed temperature profile is propagated through the bed = 191F/min.
with a velocity Wg/(Gg + Gs), but, in addition, there is ,a continual increase in Figure 12.10 illustrates the results. The maximum temperature reached is
temperature. The rise is at a rate of QR/( Gs + Gg), deg/min, or, if we plot the tem-
perature as a function of distance down the bed, we give the thermal gradient in
Tg0 +( QR
Gg +Gs
)tf = 1800 + (191)(2.36)
the heat-affected zone of the bed by = 2239F.
tan rx = QR/(Gg + Gs). This temperature rise can have disastrous consequences if the maximum tempera-
ture exceeds the melting point of the material being processed (Fe 2 0 3 in this case),
For example, in Fig. 12.9, a bed of solids is heated for a short time, 0 < t < t 0 , or is so high as to damage the equipment.*
with a hot, inert gas, and then, at time t 0 , the incoming gas changes to a reactive
gas at a lower temperature T,.. The profile is propagated as in Fig. 12.8, except ~ 2 \82_!'.--2239F
\99\0f ----r
that now, as time progresses, the temperatures at all points increase proportionally .,,~
1[ (Wz)
(.)
;::l
"O "O
() = (Jg =, es = T0 -- yos0 = - 1 - erf </> + exp -kg . erfc </> ' (12.26) ~"' 0.4 ~"'
Tg Ts 2 eff I
0.2 I
where \
\~
~-
0.0
VYgGgt 0 2 4 0.0 0 2 4 6
</>= 4G.k.rr - z 4k:;;t V (a)
z. ft
(b)
z. ft
Elliott 4 has computed a useful graph for application of this equation to Fig. 12.12 Reduced temperature graph for the shape of thermal wave when ~/G, is 0.4 ft/min
modify the shape of a thermal wave moving within the bed, and we present it and kerr is 0.01 Btu/min-ft F.
in Fig. 12.11. The application of this graph is as follows. Suppose that a cold bed,
8s = 0, is subjected to a hot gas, (Jg = 1. Since we are assuming an infinite heat- <!:> t = 100 min
transfer coefficient, the profile would be propagated down the bed as in Section
~
i::
;::l
1.0
t=0
r--1
12.4. If the conditions shown in Fig. 12.12 are used, after 9 min the pulse looks as
"'a.E 0.8 I I kerr= 0.0
shown by the solid line in Fig. 12.12(b). To see how conduction would affect the I
I
1.---J
I
2
profile, we now refer to Fig. 12.11, and compute the distance term for various "O 0.6 I I
distances around the leading edge (e.g., 2 < z < 4.5 ft, t = 9 min, VYg/Gs = 0.4 "'
(.)
;::l
"O 0.4
ft/min, and an assumed k.rr = 0.01 Btu/min-ft F). Using the curve for the leading ~"'
0.2
edge of the pulse, we find the reduced temperature in the bed which is shown in T"g
Fig. 12.12. Figure 12.i3 shows a similar result for an initial pulse of hot gas 0.0
0 2 4 6 8
followed by cold gas. z, ft z, ft
The value of k.rr used in Fig. 12.12 is reasonable for many oxide materials. (a) (b)
We could hardly expect a higher value, although it could be lower. Thus it becomes Fig. 12.13 Graph of thermal wave for ~/Gs = 0.4 ft/min and kerr = 0.1 Btu/min-ft F. The
incoming gas is reduced to the original low temperature after a short initial period of heating.
4 (---indicates profile if kerr = 0.0 instead of0.1.)
J. F. Elliott, Trans. AIME 227, 806 (1963).
l
,-
---------------------,-.,--..----1~1----------.-ne-enecroi-a-nn1re-n-e-at=trab-srerc()-emc1ettt=stanonary
obvious that the effect of thermal conductivity in spreading the pulse is usually Example 12.5 A bed of granular material, 1 ft deep, must be heated from 70 to
Ded--'123
small, but its effect on the maximum temperature reached by a pulse may be 1800F.
sizeable, and should be considered when looking at the temperature history of
material in a packed bed. a) Calculate the time required to accomplish this under the assumed conditions:
~ = 10.0 Btu/min-ft 2 F,
12.7 THE EFFECT OF A FINITE HEAT-TRANSFER COEFFICIENT- Gs = 20.0 Btu/ft 3 F,
STATIONARY BED hS = 50.0 Btu/min-ft 3 F,
If we remove the condition that hS is infinite, so that T, -=F 'I'g at all z and t, and if V0 g = 500 ft/min,
we continue to assume that the thermal conductivity of the solid is infinite, and Ts0 = 70F,
. ,, again consider the case of no reaction taking place, then we must deal with the T~ = 1900F.
" , solution of the entire basic energy equations as written with QR = 0. We can obtain b) Prepare a graph of the temperature profile of the gas and solid after 3 min.
an analytical solution, 5 but it is in the form of Bessel functions and is not easy
to use, except when carrying out computations on a computer. Solution
However, if we define new variables a) Calculate Y, with z = 1 ft.
hSz hSz 50 x 1
Y=-
Y- - -50
..
~ -~-10:0-
and
hS
Gs
z=
wz),
Vog
(t _ The reduced temperature required is
a graphical solutioH to the energy equations is available, Fig. 12.14. We describe () _ T, - Ts0 _ 1800 - 70 _
s - yo _ yo - _ - 0.945.
the use of this graph in the following example. g s 1900 70
g
To
O~-----------lr------~
.:::
rtg
1000
12 rt 1000 2000 3000
g
Temperature, F
(a)
To
g
0 0.0 0.5 1.0 = 0
,;
u
"'"'
Figure-Example 12.5: Solid and gas temperatures are given as functions of distance after ;;;
'-
3min. i:S
Figure 12.15 indicates experimental temperature profiles 6 from sintering /'"
research studies, and the photographs show graphically how sh,arp the flame front
can be. In this case, the gases are drawn down through the bed by suction. A
small amount of coke, 6 %, is mixed with the ore and burns, providing the necessary 12
rtg 1000 2000 3000
continuation of the flame front; in fact, the temperature increases slightly. As the
Temperature, F
front progresses, it spreads out, primarily because 9f the heat-transfer coefficient (b)
6
R. A. Limons and H. M. Kraner, paper in Agglomeration, Interscience-Wiley, 1962.
7
C. G. Thomas, article in Energy Management in Iron and Steelworks, Special Publication Fig. 12.15 Photographs of iron-ore sintering process in a laboratory study. Unsintered
No. 105, Iron and Steel Inst., London, 1968, page 87. mixture of ore and coke is placed in glass tube, ignited (top photo), and then after 30 sec,
the cold air is drawn down through the bed until the flame front reaches the bottom. (From
R. A. Limons and H. M. Kraner, ibid.)
12.4 Calculate the time to cool a bed 18 in. deep, which has uniformly sized t-in. diameter
spheres (w = 0.4), from a uniform initial temperature of 2400F to a maximum temperature of
200F. The airflow rate is 100 ft 3 /min per ft2 of bed surface, and the entering gas temperature
is 70F; p s = 250 lb/ft 3 .
\1
20
2cici~ci3
V0 gtCg
Le,( I - w) 0 68
PROBLEMS
' 1 *By underlining carbon we have specified its state ijust as (g) means gaseous state). In this
instance,~ means that carbon is in solution in the iron.
429
---------------------------
The rate at which the iron carburizes is related to the system variables, such as the
gas flow rate, gas composition, pressure, temperature, sheet thickness, etc. The
following consecutive steps are necessarily involved in the process:
1. Transport of methane (carbon source) in the direction of the gas flow. 13
2. Transport of methane normai to the surface of the iron.
3. The phase boundary reaction, possibly occurring in several steps. FICK'S LAW AND DIFFUSIVITY
4. Diffusion of carbon from the surface into the interior of the iron sheet.
5. Counter-diffusion of hydrogen gas back into the gas stream.
OF MATERIALS
Step (1) involves supplying the necessary methane to the sample by bulk gas
flowing into the tube and past the sample. When the gas stream flows past the iron
sheet, a velocity profile is established, and if the reaction proceeds, the methane is
depleted at and near the solid-gas boundary; thus step (2) involves diffusion with Transport of mass by diffusion is complex. When considering diffusion processes
convection. Then the methane having reached the iron surface decomposes there, in multi-component situations, we must realize that the rates of movement of the
in some manner, with carbon atoms entering the lattice-step (3)-and finally- various components can be different from each other, and that the rates to a large
step 4-carbon diffuses into the bulk of the iron. In step (5), the gaseous reaction extent depend on the nature of other components present and on their concentra-
product, H 2 , must diffuse back out into the b~lk gas stream, in order that the tions. With this in mind, let us consider first the rate equations for diffusion, and
reaction may proceed. then discuss the values of the diffusivity and the prediction of these values in various
In general, most metallurgical processes involve such a variety of different steps materials.
occurring in gaseous, solid, or liquid phases, and at different phase boundaries.
When one step is much-slower than the others, a limiting condition is reached and
this rate determining step determines the overall reaction rate; if none of the steps is 13.1 DEFINITION OF FLUXES-FICK'S FIRST LAW
much slower than the others, a more general condition-mixed control-is
Diffusion is the movement of a species from a region of high concentration to a
achieved. This parallels heat flow problems in that the overall diffusion of heat into
or out of a body may be controlled entirely by its thermal diffusivity or by the heat- region of low concentration; in general, the rate of diffusion is proportional to the
transfer coefficient at its surface, depending on their relative magnitudes. By way of concentration gradient. Consider Fig. 13.1, which depicts a thin plate of pure iron.
review, a good example of this is the cooling of solids which, as limiting conditions,
either obey Newton's law of cooling, or, if the heat-transfer coefficient approaches x=
infinity, cool at a rate entirely controlled by the solid's ability to diffuse heat.
In Chapter 13 we present Fick's law of diffusion and deal with various types of
diffusion coefficients, relating them to each other. Then we discuss their prediction
in various phases and conditions of applicability. Chapter 14 contains a variety of
solutions to Fick's first and second laws under diffusion-controlled conditions
which are often encountered in metallurgical situations in the solid state. Chapters x=L
15 and 16 provide an introduction to mass transfer in fluid systems and the com- Fig. 13.1 Establishment of steady-state concentration gradient; Fick's first law.
bination. of chemical kinetics with mass transfer.
Both sides of the iron plate are.exposed to the same pressure of hydrogen which
means that after equilibration, the concentration of hydrogen dissolved in the iron
is fixed and uniform across the plate. At some instant, t = 0, the upper surface is
subjected to a much higher pressure of the gas, which establishes a new hydrogen
concentration at that surface. The material beneath this surface is gradually
enriched as the hydrogen diffuses from the high concentration at the upper surface
into the low concentration region. A steady-state concentration profile is eventually
reached when a constant rate of hydrogen mass is required from the gas phase at
431
x = 6, in order to maintain the concentration difference across the plate. This 13.2 DIFFUSION IN SOLIDS
example is, of course, analogous to Figs. 1.4 and 6.1, and may also be described by
a rate equation. Over the past decades, a large emphasis has been placed on determining diffusion
If the concentration of component A is given in mass units, then the rate mechanisms in solids, in the hope that eventually diffusion coefficients can be
equation for diffusion may be written predicted for a given set of conditions without the necessity for carrying out the
long experiments involved in their determination. But we have to admit that that
-DA(~:)
goal has not been reached. However; it behooves metallurgists to understand the
ffAx = (13.1) proposed mechanisms, and especially the implications of the terms self-diffusion,
intrinsic diffusion, interdiffusion, interstitial diffusion, vacancy diffusion, etc., so
where WAx = mass flux of A in the x-direction, g( of A)-cm - 2 sec - 1 , p A = mass con- that they can make intelligent estimates of diffusivities, and/or the effects of alloying,
centration of A, g(of A)-cm - 3 (of total material), and DA = diffusion coefficient, or for practical uses. In the following sections, we shall consider the various types of
diffusivity of A, cm 2 -sec- 1 . diffusion coefficients and some of the proposed mechanisms for diffusion.
In metallurgical applications the rate equation is usually written in terms of
molar concentrations:
13.2.1 Self-diffusion
Ifwe could stand inside the lattice of a solid, we would see a cont;inual motion of the
(13.2)
atoms, each vibrating about its normal lattice point. Furthermore, we would see
occasional unoccupied sites, that is, vacancies. If we focus on a vacancy and its
where jAx = molar flux of A in the x-direction, mol (of A)-cm - 2 sec - 1 , and CA = immediate neighboring atoms, we will eventually see the site suddenly become
molar concentration of A, mol (of A)-cm- 3 (of total material). occupied and one of its neighboring sites become vacant. In this way, a particular
Equations (13.1), (13.2), and other altern!tive statements of the rate equation atom can slowly progress through the lattice. Another way to think of it is to
are all referred to as Fick'sfirst law ofdiffusion which states that species A diffuses in consider that the vacancy wanders randomly through the lattice. At any rate, the
the direction of decreasing concentration of A, similarly as heat flows by conduction net effect is a random motion of the atoms themselves.
in the direction of decreasing temperature, and momentum is transferred in viscous The rate at which an atom meanders through the lattice of a pure metal is the
flow in the direction of decreasing velocity. self-diffusion rate. We can measure it using radioactive atoms (tracers), as illustrated
Equations (13.1) arid (13.2) are convenient forms of the rate equation when the in Fig. 13.2, which depicts self-diffusion between a central region initially with a
density of the total solution is uniform, that is, in solid or liquid solutions, or in a uniform concentration CJ of radioactive atoms and two adjoining regions initially
dilute gaseous mixture. When the density is not uniform, other forms ofFick's first containing only normal atoms. It is assumed that the diffusion behavior of normal
law may be preferable. For example, the mass flux may be given by and of radioactive atoms is virtually the same, and by using a scheme as in Fig. 13.2
ap;) 4- "'
o E 0000000000
WAX= -pDA ( ox , (13.3) "'0 000000000
where pis the density of the entire solution, g-cm - 3 , and pJ is the mass fraction of A.
-
- "'"'
o-
E~
ooooooeooe
ooeooooeoo
0000000000
oooooeoooo
"' u
"'"'
u 0 ooeoooooeo
"'- ooooeooooo
Similarly, the molar flu~ may take the form o"O
u i:: oeooooeooo
oooooooeoo
000000000
oooooeoooo
(13.4)
in which C is the local molar concentration in the solution at the point where the 0 t=O t=t x 0
~ t = 00 x
gradient is measured, mol (of all components)-cm- 3 (of total solution), and XA is (a) (b) (c)
the mole fraction of A (CA/C) in the solution. 0 A normal atom
The diffusivities DA, appearing in Eqs. (13.1)-(13.4), are of a particular type, eA radioactive atom
known as the intrinsic diffusion coefficients, since they are defined in the presence of Fig. 13.2 Self-diffusion in materials. (a) Condition before diffusion. (b) Condition after some
a concentration gradient of A. diffusion. (c) Uniform condition after prolonged diffusion.
I.
-~---~- ....---- ....- ..... "'_"" .......""",]-.,.--........".-......,- --------,A7ru_,.V1-l~Vllt.r.)-t:';:J-----
we can measure the rate at which the atoms diffuse among themselves, that is, self-
diffusion.
We can also speak of self-diffusion rates in homogeneous alloys in the same
manner as we do for pure metals. In an alloy of gold and nickel, for example, we
can determine the self-diffusion coefficient of gold atoms D1u, if the central layer
of the homogeneous alloy initially contains radioactive gold atoms in the same
proportion to Ni atoms as in the outer layers (Fig. 13.3). Similarly, if radioactive
nickel atoms are initially present in the central layer, then we can determine its
self-diffusion coefficient, D~i. Figure 13.4 gives self-diffusion data for the entire
range of compositions in the gold-nickel system. It should be emphasized that
," q, self-diffusion data apply to homogeneous alloys in which there is no gradient in
' chemical composition. Also, Df values in an alloy depend on the composition of the
alloy, and the temperature.
Fig. 13.4 Diffusion coefficients in gold-nickel alloys. 15 is the interdiffusion coefficient and
QeOOOIO<li' 00000 1000@1000 D* the self-diffusion coefficient. (From J.E. Reynolds, B. L.Averbach, and M. Cohen, Acta Met.
Oll*IOOelO f&Ooeooo oeooeo
@aeoeo1i!llio oeooOP@@ ooeooJo 5, 29 (1957).)
(.)
01oeo!!0 o OOi&Oeeoo l\SOOOJOe
OOl&l;OeOO<&
1 OeeoOiiWO OJJoeoeo ~
"11doollllo oio oeoo'i! eOJOOlO NE ~
O!IU!IOOt!JOloo JOO l OOG!i ooo.eooo (.)
B Na
(a) (b) (c) * i:: In
O A atom q 10-2 "0. Ag
B atom E Sn
2
B atom (radioactive) Oil Pb
- Ga
Fig. 13.3 Self-diffusion in an alloy. l0-4 OJ
;:;;
Hg
You may recall from Chapter 1 that Einstein proposed the equation
10-
(1.17)
to describe self-diffusion, where B* is the self-diffusion "mobility", a measure of /
~-' 10-
the ability of an atom to migrate within the structure without any external field or ..........
chemical free energy gradient providing a driving force. /
The probabilistic approach has shown that the self-diffusion coefficient in a 10-10
simple cubic lattice may be expressed as 1
(13.5) 10-12 /""
where [J is theinteratomic spacing, and vis the jump frequency. From measurements
of D* and b, the jump frequency appears to be approximately 108 -10 10 sec- 1 ,
10-14
which is 1 in every 104 or 10 5 vibrations per atom. Most studies of self-diffusion
are concerned with this sort of information, that is, with the structure and motion
2.0 l.8 l.6 l.4 l.2 l.O 0.8 0.6 0.4 0.2 0
1
See, for example, P. G. Shewmon, Diffusion in Solids, McGraw-Hill, New York, 1963, or TM/T
R. E. Reed-Hill, Physical Metallurgy Principles, Van Nostrand, Princeton, New Jersey, 1964, Fig. 13.5 Self-diffusion coefficients for pure metals, plotted as logD versus 'JM/T, where 'JM is
pages 255-256. the melting temperature. (From 0. D. Sherby and M. T. Simnad, Trans. ASM 54, 227 (1961).)
l Mo in Fe results of tracer studies, is thus slightly lower than the true self-diffusion coefficient
Bee 2 Co in Fe D(. They are related by means of the correlation coefficient, f:
{ 3 U in U + l 0 at.% Zr
4 U in U + 18 at. % Ti
D[ =JD(. (13.6)
5 abcde Ni in Ni-Au (100,80,65,20,0 at.% Ni)
6 abed AuinAu-Ag (l00,75,25,0at.%Au) Table 13.1 gives f calculated for several different crystal structures. For interst.itial
7 abe Ti, Al, W in Ni
8 a Fe in Fe+ 2.5 at. % C diffusion, f is equal to 1.0, since the host lattice is not involved in the mechanism.
Fee b Fe in Fe+ 6 at. % C
9 a Cu in Cu + I at. % Zn Table 13.1 Correlation coefficients
b Zn in Cu + l at. % Zn
10 Ag in Ag-Pd for various structures*
(0 to 22 at % Pd)
Structure f
. ~ 1l a B or P in Si
_ Diamond b Tl or In in Si
10 9 type e Ga in Si Diamond 0.500
d As in Si
Simple cubic 0.655
Body-centered cubic 0.721
Face-centered cubic 0.781
10-11
..
- - - - - - - - - - , : a 7 1 1 ... "Uo:!'nlw:l-lll~IU..-:9--~-----
mechanisms by which substitutional alloy elements are thought to diffuse. More medium, and thus contributes to the velocity of the medium. Refer to Fig. 13.7,
detailed discussions of diffusion mechanisms in solids are available in most texts on for example, and consider a diffusion couple made by joining a gold and a nickel bar
materials science or physical metallurgy. together so that there is diffusion across the marked interface. The inert markers
Vacancy mechanism. This is a mechanism by which an atom on a site adjacent may be pieces of fine tungsten wire (insoluble in the alloy) which are located in the
to a vacancy jumps into the vacancy. While some distortion of the lattice is required plane of joining. During a diffusion anneal of many hours at a sufficiently high
for the atom to pass between neighboring atoms, the energy associated with this temperature, such as 900C, intt<rdiffusion of the gold and nickel occurs and
distortion is not prohibitive, and this mechanism is well established as the pre- changes the concentration distribution as shown. But, because gold diffuses more
dominant one in many metals and ionic compounds. rapidly than nickel, more gold atoms than nickel atoms diffuse past the inert
Ring mechanism. In some bee metals, it is thought there might exist a mechanism markers. Figure 13.7(b) shows in a very exaggerated schematic form the transfer of
of diffusion in which a ring of three atoms may rotate, resulting in diffusion. This gold atoms across the plane of markers without any nickel atoms crossing over. If
possibility is considered to be more plausible than an exchange of two atoms, since the vacancy concentration remains uniform, that is, the volume is constant, this
it involves less strain energy than a two-atom exchange. However, direct evidence transfer of gold atoms requires that the bar of pure gold must shorten while the
of either of these mechanisms operating in metals is lacking. nickel bar (now containing the transferred gold atoms) lengthens by the same
Interstitialcy mechanism. This mechanism involves the addition of an extra amount. In the same manner, the transfer of nickel atoms across the plane of
solute atom to the lattice, by pushing an adjacent atom out of its normal site and markers without any gold atoms crossing over is shown, producing the shift due to
into an interstitial site. The motion continues as the new, oversized interstitial atom Ni. The net result of these two processes, which occur simultaneously, is that the
pushes a further atom out of its normal site in a chain-reaction type of process. side that was originally pure gold is somewhat shorter, and the other side is
This mechanism is believed to operate in some compounds in which one atom is correspondingly longer. An alternative description of the phenomenon could be
smaller than the other, for example, in AgBr, where silver diffuses via this mechan- that the inert markers have moved from their original position toward the gold end
ism. In most metals, however, this mechanism seems unlikely to operate. We find of the specimen. This is the viewpoint taken if one end of the specimen is the
a possible exception in materials subjected to bombardment by radiation, in which reference plane, and the movement is called the Kirkendall shift or the Kirkendall
case high energy particles may knock atoms out of their normal sites and into effect. The differences between the various "types" of diffusion coefficients men-
interstitial positions. ~ tioned above are related to the Kirkendall effect, and are pointed out below.
Thinking of diffusion in general, we can see now the basic difference between Due to bulk motion in the gold-nickel diffusion couple, we note that the solid
diffusion and heat flow quite clearly. Heat flow taking place in a medium does not bar moves with a velocity depending on the difference between the rates of diffusion
cause the medium to move while diffusion, in itself, involves the movement of the of gold and nickel. An observer sitting on a lattice plane moving with the solid's
velocity would notice a flux of gold atoms past that plane given by
. oCAu
lAu = -DAu~
0
However, an observer sitting on an unattached plane in space, would see the total
flux of gold atoms passing by as
(13.7)
(c)
where the flux is expressed by NAu rather than by jAu to emphasize that it is
relative to the stationary plane. The accumulation of gold as a function of time
Net
within a unit volume that straddles the stationary plane is given by the net difference
shift between the gold entering and leaving. Thus, referring to Fig. 13.8, we may write
(d) for a unit area perpendicular to the page :
x
Fig. 13.7 The interdiffusion of gold and nickel, and the resulting bulk flow and composition (13.8)
profiles.
~-~ -..----r..,:-.:----------------------D11111S1on-nrsonas------zizn-----
this requires that the position of a reference plane fixed in space can be determined.
f..-----!- Stationary plane
I
I
In the experimental set-up of Fig. 13.7, the inert markers could provide such a
I
I reference plane. But in applying diffusion data to a commercial process, we cannot
expect to be provided with inert markers. In addition, Eq. (13.15) is useful for
b.x determining DAu and DNi, but is of no value in the sense that analytical solutions to
I
' practical diffusion problems cannot .be conveniently developed from it. Instead,
x x + b.x the basic complexity that arises from the bulk motion is avoided by merely using
Fig. 13.8 Unit element for mass balance. Fick's first law in its simplest form, that is, for gold
oCAu oNAu Note that the diffusion coefficient is written differently; D includes the basic
(13.9)
ot -a-;:-' concept that the flux is proportional to the concentration gradient but side-steps
which, on substituting Eq. (13.7), becomes the issue of bulk flow. With D defined in this way, the accumulation of gold within
the unit volume of Fig. 13.8 is
OCAu 0 ( oCAu )
-at = ox DAU a-;- - (13.10)
L\x( 0 ~;u)
vxCAu .
= UAulx - jAulx+!ixJ. (13.17)
Similarly, we can obtain the accumulation of nickel in the same unit volume: When taking the lim , we obtain the unsteady-state equation for unidirectional
&x-+O
diffusion in solids:
(13.11)
If the vacancy concentration within the unit volume is constant (volume constant),
oCAu = _!!_
ot ox
(jj oCAu).
ox
(13.18)
then
We use this equation, which is often called Fick's second law, to obtain analytical
oC oCNi oCAu solutions for diffusion problems. By comparing Eqs. (13.15) and (13.18), the
-=-+--=0 (13.12) relation between the more fundamental quantities, DAu and DNi-called intrinsic
ot ot ot '
diffusion coefficients-and the more useful quantity, D-called the interdiffusion,
which requires, by Eqs. (13.10) and (13.11), that mutual diffusion, or chemical diffusion coefficient*-is evident
where nA is the number of A atoms per unit volume. Recalling (from thermo-
( ~)
aln XA =1
'
dynamics) that and
(13.30)
then known as the Nernst-Einstein equation. If the mobility BA, in the presence of a
driving force, is independent of composition, then DA in an ideal solution alloy is
(13.24)
independent of composition. However, it is reasonable to expect BA to be a function
of composition in alloys and non-stoichiometric compounds and so, even if the
Finally, by combining Eqs. (13.23) and (13.24), we can write the flux of A in terms solutions are thermodynamically ideal, DA may still be a function of composition.
of the mobility, BA, and activity, aA, as follows: In addition, due to the fact that DJ = BJKs T for self-diffusion, it is often stated
. - -nABART
nAx - No
(a lnaxaA) (13.25)
that Eq. (13.30) may be written as DJ = BA Ks T for an ideal solution. However,
this is not true, unless fortuitously Bl = BA, for there is no reason to think that BA
under the influence of an energy gradient should be the same as BJ with no energy
Now comparing Eq. (13.25) to Fick's first law (Eq. 13.2), we find that gradient present. Figure 13.4 shows sufficiently clearly that D~ is not a constant
with composition; at best, one would have to know Bl as a function of composition,
DA(a1;xcA) = BAKsT(a~nxaA), (13.26) and since B values are usually computed from diffusion coefficient measurements,
& this becomes an exercise in rhetoric.
with Ks ('Boltzmann's constant)= R/N0 = 1.38 x 10- 23 joules-deg- 1. Only in one instance-in nonmetallic stoichiometric compounds-may BA
and BA be the same, and DA be calculated from an independent measurement
Since we can express the molar volume of A (cm 3 rriol- 1 ) as XA/CA, then
(such as ionic electrical conductivity) of BA. We shall learn more about this subject
acA _ cA (axA) in Section 13.3.
ax - XA ax ' Returning to Eq. (13.19), we can modify the expression. From the Gibbs-Duhem
equation, we know that
*A self-consistent set of units for use with this and following equations relating electrical and
diffusive properties is G; = joules-mo!- 1 ; B; = cm 2 -sec- 1 joules- 1 ; K 8 = joules-deg- 1 ; (13.31)
n; = ions-cm- 3 ; e =coulombs;;= joules-atoms- 1 .
----~-~..I-;:>;-.<;
---------------1l7JITUSIUD-in-:su11~':'.'
t = 10 hr
so that in this case, if we know the self-diffusion coefficients, Di and DJ, and the
way in which the activity of A varies with composition, we can calculate the inter- (b)
. " diffusion coefficient, which is the useful value for engineering calculations. The
thermodynamic term, which is sometimes called the thermodynamic factor, can t=IOOhr
also be expressed in the form
(c)
(13.35) I =oo
But since the restriction that BA equals BJ, and BB equals Bf, still operates, Eqs.
(13.35) and (13.33) are of limited application. However, Eq. (13.32) gives some Concentration profiles (e) Free energy profiles
insight into the conditions where diffusion may occur from the point of view
of a concentration gradient. The best-known example is the experiment carried ---%Si - - - Gsi
--3c - - Ge
out by Darken. 3 He welded two pieces of carbon steel together and studied the
diffusion of carbon, as schematically shown in Fig. 13.9. One of the bars contained
3.80 % Si and 0.48 % C, while the other contained only 0.44 % C. Figure 13.9(a) Fig. 13.9 Concentration and partial molar free-energy curves at various times for Fe-C and
Fe-C-Si alloys welded together and annealed.
shows the initial carbon distribution and we might expect that the carbon distribu-
tion would simply have evened out at 0.46 % C. However, because silicon greatly
increases the activity of carbon in iron, the carbon in the left-hand bar was at a
much higher chemical potential than it would have been at if no silicon had been 13.2.4 Temperature dependence of diffusion in solids
present. The result was that a sizeable amount of carbon diffused from left to right,
and after a short time, the carbon gradient was as shown in Fig. 13.9(b). If the Temperature has a tremendous influence on diffusion in soli~s. It has em~irical~y
silicon did not diffuse, the final profile would look as in Fig. 13.9(c). If diffusion been found that the Arrhenius equation adequately descnbes the relat1onsh1p
continues until both chemical potential curves are flat, not only for carbon but & between any type of diffusion coefficient and temperature:
also for silicon, then the silicon will eventually diffuse to the right-hand side. D = D0 e-QfRT, (13.36)
The diffusion of silicon will be much slower than that of carbon, since it involves
substitutional diffusion rather than the interstitial mechanism by which carbon in which Q is the activation energy, and D0 is sometimes called the frequency factor,
diffuses. This means that the higher carbon concentration (created on the right- both essentially constant over a wide temperature range.
hand side before appreciable silicon diffusion occurred) will then decrease until Much effort has gone into the development of theoretical expressions for Do.
ultimately both the silicon and carbon gradients become flat (Fig. 13.9e). Recall that the self-diffusion coefficient D* was expressed earlier in terms of the
jump frequency v according to
>
3
L. S. Darken, Trans. AIME, Inst. Met. Div., Tech. Pub!. 2443 (1948). (13.5)
L
T,C
,
1
o- 950 900 800 100 650 600 550 500475450425400 0
0
0
0
0
0
0
0 0 0 0 T, C
~ Si (<;;;2.5%) -- -
M N 0 0
a..
0
00
0
r-
'. )
a ''
'\.'\ '
I
u
IQ
' -.
'-
'"'\ !. ' '\.
u
~
. I I
\\ \ ~\. \. '\ ' C in Fe+ 2.5%Cr
\ ' \ \
\\
. ~\ \\. I\ .
' \ \
'\
'I\
' ..
\
Zn(20-25%)
' \ \''\ \ ,zn-
'\
.
\
\ \ \ \. \. '\ '\ \
\ ~,\ '.i I
Ag(l.2%)- ~' '\. \ \ \
,\
\ \~, \ I~ \
\ C in (Fe+ 1 %Mn)
\also C in (Fe + Ni)
10-10
.
\
. Cu(<;;;35%l
. . '\ \
10-11
\ \
~
~ \ C in Fe
I \
'\ "
10-12
"
. Ni(7 .5-11.8%
I
'.
''
r ~.
I\
\
..' ..
.,,
''
'\
':'
,
\
\A1(15-20%)
\
\
\
10-7
I. I "" '
1.2
lx1030K-1
T
"'
'
1.3
l_
0
0
M
0
0
N
0
0
0
0
0
0
0
'
0
0
00
0
0
r--
0
0
'
T. C
0
0
V)
r
I
it also applies to intrinsic diffusivities of solutes in dilute concentrations in a variety
of solvent hosts, with reasonably satisfactory results.
hep 17
Diamond 21 *Jost, Diffusion, Academic Press, New York, 1960.
K. Hauffe, Oxidation of Metals, Plenum Press, New York, 1965.
For estimation purposes, D0 is approximated as 1 cm 2 /sec, as shown for most 0. Kubachewski and Hopkins, Oxidation of Meta ls and Alloys, Plenum Press, New York, 1962.
metals in Fig. 13.5. Furthermore, Q is predicted by P. Kofstad, High Temperature Oxidation of Metals, Wiley, New York, 1966.
F. A. Kroger, Imperfections in Crystals, North-Holland, New Amsterdam, 1964.
(13.41)
t The notation [l::t ] refers to the concentration, cm - 3 , of vacant A sites, [A;] refers to the
This correlation has been tested for self-diffusion in alloys, in which case TM is concentration of A atoms on interstitial sites, and [A 8 ] to the concentration of A atoms on sites
taken as the temperature halfway between the liquid us and solidus for the alloy; nonr{ally occupied by B atoms.
__T _ _
1
with charge neutrality maintained by the creation of Fe 3 + ions according to the The electrical conductivity a is defined by*
equation
3Fe 2 + ---> VFe + 2Fe 3 + + Fe 0 (removed). (13.46)
Some of the trivalent iron ions may be ionized according to*
Fe 3 +---> Fe 2 + + EB, Now recalling Eq. (13.30)
which results in an increase in electronic conduction via electron holes as further
deviations from stoichiometry occur.
we can relate the ele~trical conductivity and diffusivity by the simple expression:
Because of the electrical nature of the defects and of the diffusing species (ions),
ionic electrical conduction and diffusion are inseparable processes. Consider an
ionic material in which one ionic species makes the predominant contribution to D; = K T '
(13.47)
8
electrical conduction (for example, Na+ cations in NaCl). The general equation
for the transport of the species in the x-direction under the influence of an electric whereK 8 = 1.38 x 10- 23 joules-deg- 1 . Thisequationisknownasthe"extended"
field and a concentration gradient in the x-direction is: Nernst-Einstein equation.
In order to predict D; from conductivity measurements, we must know the
fraction of the total conductivity due to species i; this fraction is called the trans-
fl.
IX
= -v.(on;)
IOX - B.n.z.e( ox<P),
0
I I I ference number of species i, denoted t;. This number, obtained from electrolysis
experiments, gives the fraction of the total current carried by a particular species,
where z; = number of charges on the diffusing species, and the total of all transference numbers must equal one. Therefore
n; = concentration of diffusing species, ions-cm - 3 ,
B; = mobility of species i (steady-state velocity of the particle under the
L
cation
tcationi + L
anion
tanioni + L telectron = 1. (13.48)
influence of a unit force), cm 2-sec- 1 -volC 1 -coulomb-1, species species
* Note that in the physics literature, a = ne, where is called the mobility. Since this is
applied to electron or electron-hole conduction, z = -1 or + 1, and therefore, = Be. In
this case, the units of a are ohm- 1 cm- 1 , n are cm- 3 , e = 1.6 x 10- 19 coulombs, and=
* The notation EB is used to represent an electron hole in the valence band, contributing to
p-type, or hole conduction; similarly, e is the symbol for electrons in the conduction band. 6
.
cm 2 sec- 1 volt- 1 .
D. Mapother, H. N. Crooks, and R. Mauer, J. Chem. Phys. 18, 1231 (1950).
NS
=
u
"
u
1l_
-;::;
E
10- 700
10-
600 500 400 350
oc
' Presumably, diffusion of Co++ ions should be via a vacancy ~echanism .on
the cation sublattice with diffusivity increasing as the concentrat10n of cat10n
vacancies increases. The phase diagram (Fig. 13.14) shows the Co-CoO equilibrium.
When P. is increased above that existing at this lower boundary of the CoO phase
field th~:nolar concentration of Co decreases as oxygen is added to the oxide, until
finally an oxygen pressure is reached at which Xe0 = 0.492 (corresponding to
"0u Co O) beyond which Co 30 4 forms. This occurs at P02 = 0.53 atm at 1000C.
10-10 0.969 1b .
" Price and Wagner measure chemical diffusion coefficients in CoO by eqm.1 ratmg
-~ 2
e0 10-ll
a single crystal of CoO with an atmosphere at low P02 (4.7 x 10- atm); then, at
time zero, they change the oxygen potential to a higher level, near the CoO-Co304
~
boundary. The change in X 0 ( = - Xe 0 ) is measured as a function of time by
following the change in electrical conductivity, since the conductivity i~ th~s
10-12
material is proportional to the concentration of electron holes [ E8 ], which is
dependent, in turn, on the oxygen content through the equilibrium constant for the
10-13 reaction shown above;
0 0 0 0 0 0 0 0
0 r--:
- "' ~ '<!: ""!
"" K = [E8][Vo+J.
l/K X 10 3 P6; 2
Fig. 13.13 Log D versus 1/T for sodium in NaCl as determined with radioactive sodium ( o ), Although i5 is generally composition dependent in non-stoichiom!t.ric crystals, in
and as calculated from the conductivity ( ). (From D. Ma pother, H. N. Crooks, and R. Mauer,
J. Chem. Phys. 18, 1231 (1950).) this case the deviations from non-stoichiometry are small, and D is found to be
essentially constant with composition:
values is excellent at the higher temperatures. The discrepancy at the lower
temperatures points out the effect of CdC1 2 present as an impurity in NaCl in dilute
i5 = 4.33 x 10- 3 exp (-24,000/RT).
concentration. By adding CdC1 2 , [VNaJ increases, thereby increasing DNa over that Since the cobalt cation is much smaller than the oxygen anion, De 0 is very much
in the pure material. However, the conductivity expression does not include this
Table 13.2 Diffusivities in nonmetallic crystals
added effect, and the values differ when the concentration of vacancies due to
CdC1 2 is significant, compared to the intrinsic concentration of vacancies. Crystal in which
Non-stoichiometry, resulting in vacancies on the cation sublattice, can also Diffusing diffusion takes Do, Q,
have an effect on D[ (and through Eqs. (13.33) or (13.35) on D) in the same manner ion place cm 2 /sec cal/mo!
as impurities, as illustrated in the following Example.
Ag+ a-Cu 2 S 38 x 10-s 4,570
Example 13.1 Given the chemical diffusion coefficient i5 as determined from cu+ a-Ag 2 S 12 x 10- 5 3,180
electrical conductivity measurements and the thermodynamics of the CoxO Ag+ a-Cu 2 Te 2.4 20,860
system, calculate the diffusion coefficient of Co+ + ions and compare it with the cu+ a-Agl 16 x 10- 5 2,250
measured diffusion coefficient De 0 This is a problem studied by Price and Wagner. 7 Li+ a-Agl 50 x 10- 5 4,570
se- - a-Ag 2 S 17 x 10- 5 20,040
Solution. CoO exists as a non-stoichiometric oxide with defects on the cation Pb++ PbC1 2 7.8 35,800
sublattice in the form of vacancies. This amounts to the addition of an oxygen Pb++ Pbl 2 10.6 30,000
anion without addition of a cation, according to the reaction o-- Fe 2 0 3 1 x 10+ 11 146,000
Fe+++ Fe 2 0 3 4 x 10+ 5 112,000
tOz(g) = Oo + Ve~ + Ee, co++ coo 2.15 x 10- 3 34,500
where 0 0 indicates an excess oxygen ion on a regular oxygen site, Ve~ a singly- Ni++ NiO 1.83 x 10- 3 45,900
ionized cation vacancy, and E8 a free electron hole. o-- NiO 1.0 x 10- 5 54,000
er+++ Cr 2 0 3 0.137 61,100
7
J.B. Price and J.B. Wagner, Jr., Zeit. f. physik. Chemie, Neue Falge 49, 3-4 (1966).
l000C where Pg 2 is the standard oxygen pressure, 1 atm. We obtain the slope of the
~ log a0 versus log X 0 plot from the data of Eror and Wagner, 8 who measure [Vc J
N 0
2
13.4 DIFFUSION IN LIQUIDS
greater than D0 in CoO. Then we may use Eq. (13.33), neglecting the term D6Xc 0
, 13.4.1 Liquid state diffusion theories
if we further assume that Eco = .82'0 in order to calculate D2'0 from the i5 data:
Hydrodynamical theory. The earliest theoretical equation for prediction of self-
diffusion in a liquid is that ofEinstein. 10 In this theory, he assumes that the diffusing
i5 ~ D* X ( oIn ac 0
)
species are non-reacting spherical particles of radius R moving through a con-
- Co o 0 In Xco
tinuous medium of viscosity 1'/ with steady-state velocity V,,. According to Eq.
We may calculate the thermodynamic factor by noting that (2.123), the force on a sphere moving at steady-state in laminar flow is
oIn ac oIn a0 )
0 ) ( F = 6nR17V00 , (2.123)
(oIn Xc = - oIn Xc
0 0
'
and that 8
N. Eror and J.B. Wagner, Jr., J. Phys. Chem. Solids 29, 1597 (1968).
ao -_(P
1 2
02 ) 9R. E. Carter and F. D. Richardson, J. Metals 200, 1244 (1954).
1
-0 '
Po2 10
A. Einstein, Z. Electrochem.11, 235 (1908).
--- ----- ----~-- --- ------.._-::y; ...,-~-
and since the definitioJJ of the mobility Bis V00 /F, we have Hole theory. The oldest structural picture of a liquid is the hole theory which
presumes the existence of holes or vacancies randomly distributed throughout the
1 liquid and providing ready diffusion paths for atoms or ions. The concentration of
B=-- (13.51)
6rcR11 these holes would have to be very great in order to account for the volume increase
Combining this result with Eq. (13.30), we get upon melting, thus resulting in much higher diffusion rates in liquids than in solids
just below the melting point. This jump in D on melting is obvious in Fig. 13.5.
D = KsT' The hole theory, however, does not result in a prediction of diffusion coefficients
(13.52)
6rcR11 by itself, although it has been used to estimate the activation energy for self-
diffusion in a liquid, by assuming that this energy is equal to that required to form
known as the Stokes-Einstein equation. This has been modified by Sutherland 11
a hollow sphere (hole) of a diameter on the order of a few angstroms.
who adopted a slightly different form of the drag force, with the result that '
Eyring theory. Eyring et al. 13 have applied their activated state theory of diffusion,
D = KsT. (13.53) which works reasonably well in solids, to liquid diffusion, assuming that the
4rcR11 migrating atoms move from hole to hole in the liquid by a process of discrete
Althoug~ the use ?f Stokes' law for this purpose may seem unjustified, since it jumps. If the liquid is considered quasi-crystalline, and the atoms are in a cubic
. negl~cts mteratomic forces and considers only form drag, the results, in terms of configuration, then an expression relating D* and Y/ results:
predicted data, are remarkable. For example, in liquid metals, if we substitute the
measured values of D and Yf into Eq. (13.53) and then calculate the radii of the
D* = KsT. (13.54)
2R11
diffusin~ sp~cies, we obtain very good agreement between these predictyd radii and
the radn of 10ns, as shown in Table 13.3. This expression obviously predicts D* values six times larger than that given by
14
the Sutherland equation. Subsequent revisions of this equation by Li and Chang,
Table 13.3 Comparison of calculated and measured radii for self-<;liffusion in liquid metals
and by Eyring et al. 15 have improved the agreement with the Sutherland equation,
Element Temperature range, C Sutherland model Measured radii* but the modifications suggested do not shed any more significant light on the
structure of liquids, nor are they any more useful from a prediction standpoint.
Na 110-630 L51-4.53A L57A
Hg 274-364 1.20-1.24 1.44 Fluctuation theory. Because the ideas of a well-defined "activated state" as well"as
16
In 448-1013 1.36-3.06 1.50 of discrete "holes" in the liquid have been difficult to accept, Cohen and Turnbull,
Ag 977-1397 1.37-1.37 1.34 and later, Swalin 1 7 and Reynik 18 have developed a new theory known as the
Zn 394-837 1.29-0.75 1.25 fluctuation theory. In this theory, the "extra" volume in the liquid (over that of the
Sn 575-956 1.30-2.22 1.41 solid) is distributed evenly throughout the liquid, making the average nearest
*The radii used for comparison are those reported by Pauling for single-bond metallic situation. 12
neighbor distance increase. One may imagine a diffusing atom contained in a cage
whose dimensions are constantly fluctuating. Local fluctuations in density then
The Sutherland equation is quite successful in predicting self-diffusion da;a for occasionally open up holes or openings in the cage large enough to allow the atom
a wid~ varie~y ~f other substances, including molten semiconductors, polar liquids, to diffuse out of the cage. In their approach, Cohen and Turnbull consider that a
ass~ciated l~qmds, and molten sulfur. If we assume that the atoms in the liquid critical fluctuation must occur before diffusion occurs, whereas Swalin and Reynik
are ma cubic array, then for materials generally the simple theory predicts that &
both think that a spectrum of fluctuations occurs, with cooperative motion of the
2R = CV/N 0)11 3,
and 13 H. Eyring, S. Glasstone, and K. Laidler, Theory of Rate Processes, McGraw-Hill, New York,
113 1941.
D* = Ks T (N_.o) 14 J. CM. Li and P. Chang, J. Chem-Physics 23, 518 (1955).
2rc V '
15 H. Eyring, T Ree, D. Grant, and R. Hirst, Z. Elektrochem. 64, 146-152 (1962).
where V is the molar volume. 16 M. H. Cohen and D. Turnbull, J. Chem. Physics 31, 1164 (1959).
11
W. Sutherland, Philosophical Magazine 9, 781 (1905).
17 R. A. Swalin, Acta Met. 7, 736 (1959).
12
L Pauling, Nature of the Chemical Bond, Cornell Univ. Press, Ithaca, New York, 1960.
18 R J. Reynik, Trans. AIME 245, 75 (1969).
-r..---------. ----~ ... ~~-- ..... '~-~~
neighboring atoms r..::sulting in diffusive motion of an atom. No activation energy where d = average interatomic distance in the liquid, A and R = ion radius, A, the
is required, since there is virtually no difference between an atom moving forward f evaluation of this theory relies on the use of x-ray data ford and Z, and b-values
as a result of cooperative movements and an atom not moving. The result of l-1 obtained from correlations of existing data using least squares data fitting tech-
Swalin's theory* is that D exhibits a linear dependence on temperature. niques to fit an equation of the form of Eq. (13.55). From the calculated values of
Reynik has proposed a small fluctuation model which also predicts a linear X 0 and Eq. (13.56), R-values are obtained, which agree quite closely with Pauling's
temperature dependence of D on T (K): neutral atom radii, suggesting that in liquid metals the diffusing ion core carries its
' It valence electrons with it.
D =a+ bT, (13.55) I The problem of all the theoretical approaches is that the tests of the theories
where a = 1.72 x 10 ZXi;K, b = 2.08 x 10 ZX 0 , Z = number of nearest
24 9 rely on the test of the functional relationship between D and T. For example,
neighbors, X 0 = maximum diffusive displacement due to normal vibrations, A, while Reynik can reasonably fit liquid metal diffusion data to a linear relationship,
and K = a force constant. Saxton and Sherby 20 have tested a large amount ofliquid metal data and conduded
that diffusion in liquid metals is a thermally activated process, obeying an Arrhenius
Since X 0 = d - 2R, (13.56) relationship (Eq. 13.36), which is clearly at odds with a linear relationship with T.
0
T, C The activation energies Q are apparently not large, and so, when the RT term is on
0 80000 0 0 0
00000 0 0
I 10-3 -
\0
.-00\0ll")V M
0
N 0 the same order of magnitude as Q, D is not a strong function of temperature. Thus
~ no critical tests of these models will really be possible until more accurate data
su
over much wider ranges of temperature are available. Until then, we shall have to
rely on measured values and intelligent "guestimates."
2.5%C
4.6%C 13.4.2 Liquid diffusion data
I i\:~C alloys The remarkable characteristic of diffusion in liquid metals, and in liquids in general,
is the fact that the values of D are almost all approximately the same order of
IO"""
ON .\
~
I
'
'
'-"
'- .......
-.........
....
magnitude, 10- 4 -10-s cm 2 -sec- 1 , even though their solid state properties differ
.'\
\
.\ .\ .......
~ l'\..In ....
......... Na
""1....
widely. Furthermore, the activation energies are also approximately equal, usually
in the range 1000-4000 cal/mol. Because the data are usually presented in the
\ 0~ I~ Sn "'-J Hg
~
\ Arrhenius form and since there is still no clearcut evidence against this representa-
'
Ag \ \zn ~1'J47%S~ ...... tion, we shall present the data in this section in the same way.
In -In alloys
........... Diffusion in liquid metals. Figure 13.15 gives some data for self-diffusion in pure
....... liquid metals, Fig. 13.16 for interdiffusion in common nonferrous binary alloys,
/ and Fig. 13.17 for ferrous alloys. We refer the reader to Edwards et al., 21 Yang and
Derge) 2 or Ellio.tt et al. 23 for more complete listings of the available data.
Diffusion in molten salts and slags. Again, we know so little of the detailed
structure of molten halides, sulfides, oxide,s, and silicates, that no really satisfactory
quantitative expression is available for predicting diffusivities in these materials.
However, if we refer to Chapter 1 in which we discussed the relationship between
~
'" bonding and viscosity, we realize that the same semiquantitative arguments can be
10- 20
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 H. Saxton and 0. Sherby, Trans. ASM 55, 826 (1962).
10 4 /T, K- 1 21
J.B. Edwards, E. E. Hucke, and J. 1. Martin, Metallurgical Reviews 13, No. 120, l (1968).
Fig. 13.15 Self-diffusion data in liquid metals. 22
L. Yang and G. Derge, paper in Physical Chemistry of Process Metallurgy, AIME 7, 195
(1961).
23
*As corrected by R. J. Reynik, Applied Physics Letters 9, 239 (1966). J. F. Elliott, M. Gleiser, and V. Ramachrishna, Thermochemistry for Steelmaking II, Addison-
19
R. Swalin and V. G. Leak, Acta Met. 13, 471 (1965). Wesley, 1963.
L
------:cT1nu31u1r111-1n1u1u~-~uI
T, C T. C
100 27 1600 1500 1400 1300 1200
~
~ .....
..................
1---+-...-+----t-=~Bi in Sn
Cd in Pb ""'
c (I.5%) ' 'c (2.5%)
I X 10-
Si
r---
p, .._ i--.
10-sl---..,...L----'=--~--,...L_---'
1.0 I.I 1.2 I. 1.4 1.5 - s .. ~~
Mn
Ti-~
--- ......._
10 3 /T. K- 1
Fig. 13.16 Interdiffusion coefficients in liquid nonferrous alloys.
s (0.6%) I"---_ -
Ni- r - -
used with regard to diffusion in these materials, particularly in light of the Stokes-
Einstein model of diffusion. s (0.9%)..,
Consider first the molten salts. We may describe their structure as holes within 1X 10-5 1'.. ......
a molecular assembly that has density fluctuations and short-range order, but with
the added restriction of electroneutrality, so that the cation and anion charge
densities must remain equal in any region of the melt. Diffusion coefficients
measured in molten salts are remarkably close to values observed in liquid metals,
even though there is an obvious difference in bonding. Molten salts have highly
polarized ionic bonds, in contrast to the metallic bond. Table 13.4 gives some
typical data for self-diffusion in molten salts. Note that the activation energies for
the salts are, however, larger than the activation energies for self-diffusion in
metals. 1 X 10-6
0.50 0.55 0.60 0.65 0.70
In molten slags, the picture is more complicated; from the relation between the 10 3 /T. K- 1
degree of polymerization of the silicates and their viscosity, we can postul~te that Fig. 13.17 Interdiffusion coefficients in ferrous liquid alloys ( - - - carbon-saturated;
Table 13.4 Self-diffusivities in molten saltst ------pure Fe).
Typical D* the diffusivity of oxygen, for example, is lower in the more acid slags, and increases
Temp. range, D0 x 10 ,4
Q, cal-mole- 1
value, 10apidly as the concentration of (ree oxygen ions increases near the orthosilicate
Diffusate Melt oc cm 2 -sec- 1 cm 2 -sec- 1 oc composition. On the other hand, the diffusivity of basic cations should not depend
on the slag composition to such an extent.
Na NaCl 845-916 8 4,000 14.2 x 10- 5 906 There are relatively few diffusion data available for molten slags; most of them
Cl NaCl 825-942 23 7,100 8.8 x 10-s 933 are presented in Table 13.5. The curious result, which has not yet been satisfactorily
Na NaN03 315-375 12.88 4,970 2.00 x 10-s 328
10-s explained, is that the tracer self-diffusion coefficient of oxygen in a basic slag is
N03 NaN0 3 315-375 8.97 5,083 1.26 x 328
Tl TIC! 487-577 7.4 3.89 x 10-s
greater than the self-diffusion coefficients of calcium in the same slag. It is also
4,600 502
Zn ZnBr 2 394-650 790 16,060 0.22 x 10- 5 500
greater than that of silicon and aluminum, but we may explain this on the basis that
silicon is tied up in SiO!- ions, which clearly do not diffuse as easily as the smaller
t Data from L. Yang and G. Derge, ibid. 0 2 - ions; aluminum is probably tied up in AlO~ - ions, which diffuse only a
--------~---- -- ---------- -----------------------------.-----------------
fraction more easily than the SiO!- ions. Note also that the activation energies Table 13.6 Diffusivities in common liquids at 25C
are now quite large, in correspondence to the increased difficulty of movement of
these complex molecules, presumably requiring much more complicated relative Solute Solvent Concentration D, cm 2-sec- 1
movements of nearest neighbor ions in order to affect any net movement of a
HCl Water O.lM 3.05 x 10- 5
molecule.
NaCl Water O.lM 1.48 x 10- 5
Diffusion in common liquids. Table 13.6 gives some data for diffusivities in 1.10 x 10- 5
CaC1 2 Water O.lM
aqueous and organic solvents. Again, note that they are all on the order of Water Dilute 5.0 x 10- 5
H2
10- 5 cm 2 -sec- 1 , even though there are obvious differences in bonding between Water Dilute 2.5 x 10- 5
02
these liquids and liquid metals or slags. S02 Water Dilute 1.7 x 10- 5
NH 3 Water Dilute 2.0 x 10- 5
Cl 2 Water Dilute 1.44 x 10- 5
Table 13.5 Self-diffusivities in molten silicates and sulfidest H 2S0 4 Water Dilute 1.97 x 10- 5
Na 2S04 Water O.Ql M 1.12 x 10- 5
Temp. range, D0 x 104, Q, Typical D* value, K 4Fe(CN) 6 Water O.Ql M 1.18 x 10- 5
System C cm 2-sec- 1 cal-mole- 1 cm 2-sec- 1 C Ethanol Water x = 0.05 1.13 x 10- 5
Glucose Water 0.39% 0.67 x 10- 5
Ca in Ca0-Al 20 3 -Si02 1350-1450 70,000 0.062 x 10- 5 1400 Benzene CC1 4 Dilute 1.53 x 10- 5
(40/20/40) CC1 4 Benzene Dilute 2.04 x 10-s
Ca in CaO-Al 20rSi0 2 1350-1540 70,000 0.067 x 10- 5 1400 Br 2 Benzene Dilute 2.7 x 10- 5
(39/21/40) CC1 4 Kerosene Dilute 0.96 x 10- 5
Ca in Ca0-Al 20 3 -Si0 2 1350-1450 70,000 0.067 x 10- 5 1400 C02 Ethanol Dilute 4.0 x 10- 5
(40/20/40) CC1 4 Ethanol Dilute 1.50 x 10- 5
Si in Ca0-Al 20 3 -Si0 2 1365-1460 70,000 O.Ql x 10- 5 1430 Phenol Ethanol Dilute 0.89 x 10-s
(40/20/40)
0 in Ca0-Al 20 3 -Si0 2 1370-1520 95,000 0.4 x 10- 5 1430
(40/20/40)
13.5 DIFFUSION IN GASES
Fe in Ca0-Al 20 3 -Si0 2 1500 0.24-0.31 x 10- 5
(30/15/55) Through the years, diffusion studies have been mostly concerned with measuring
Fe in Ca0-Al 20 3 -Si0 2 1500 0.21-0.50 x 10- 5 and predicting diffusion coefficients in gaseous mixtures. In gases, we are not faced
(43/22/35) with the problem of Kirkendall shift phenomena, since any tendency for more A
Fe in Fe0-Si0 2 1250-1305 40,000 9.6 x 10- 5 1275 molecules to go in one direction than B molecules in the opposite direction is
(61/39)
immediately counteracted by the fact that a pressure gradient builds up; thus in
Pin Ca0-Al 20 3 -Si0 2 1300-1500 46,600 0.2 x 10'" 5 1400
(40/21/39) gases, DA = DB= DAB = 15.t Of course, one must still superimpose the bulk flow
Fe in Fe-S (33.5 %) 1150-1238 65.7 13,600 5.22 x 10- 5 1152
term in Eq. (13.7) onto the Fick's first law expression to obtain the total flux if
Fe in Fe-S (31.0%) 1164-1234 298 17,000 6.39 x 10- 5 1180 there are any convection effects present.
Fe in Fe-S (29.0%) 1158-1254 20200 27,700 11.91 x 10- 5 1158 ~ Chemists and chemical engineers have spent a great deal of time and effort in
Fe in Fe (48.1 %)-Cu 1160-1250 1440 21,500 7.57 x 10- 5 1160 developing equations which predict diffusion coefficients in gases. For example,
(20.5 %)-S (31.9 %) based on the kinetic theory of gases, the following equation applies to prediction of
Fe in Fe (32.0 %)-Cu 1168-1244 35.7 13,700 2.94 x 10- 5 1168 self-diffusivity of spherical A atoms diffusing in pure A :
(40.0%)-S (28.0%)
Cu in Cu-S (19.8 %) 1160-1256 69.3 12,800 7.49 x 10- 5 1160 * - ~ ( I(~ )1/2 y3/2. (13.57)
Cu in Fe (32.0 %)-Cu 1160-1245 564 19,700 5.52 x 10- 5 1160 DAA - 3 n3mA Pd2
(40.0 %)-S (28.0 %)
t Chemical engineers often use the notation DAB to refer to the diffusion of A in an A-B mixture.
t Data from L. Yang and G. Derge, ibid. This is the same as our D, which, if DA = DB, is also equal to DA and to DB.
l
------.-------------- --------- -- ------------------ - - - - - - - - - - - - - - - - - - - - -
For the interdiffusivity of two unequal si:ze spherical atoms A and B, kinetic theory and
predicts
(
8
Ks AB
)
=
[(
8
)
Ks A .
(
8
Ks B
) J112
= average intermolecular force parameter, K;
(13.58) MA, MB = molecular weights of species A and B;
T = temperature, K;
P = pr~ssure, atm.
where =Boltzmann's constant, 1.38 x 10- 16 ergs-molecule- 1 K- 1 ,
K8
The values of the force parameters and the collision diameters for common gases
d = molecular diameter, cm,
are given in Table 1.1, and the collision integrals iln,AB are given in Figure 13.18,
m = molecular mass, grams-molecule-1,
from which we may calculate DAB values for these gases. Note that the c?llisi?n
P =pressure, dynes-cm- 2 ,
integral values for diffusion are only slightly different from those for gas v1scos1ty
T = temperature, K.
calculations.
Equations (13.57) and (13.58) give the proper pressure dependence up to about
The force parameters for metal vapors are not well known, but Turkdogan 25
10 atm; the predicted temperature dependence is only qualitatively correct in that
24 has estimated some of them. Remembering that Hirschfelder et al. 26 showed that
DAB actually varies more with temperature.
in the case of almost spherical molecules the reduced pressure P,, volume V,., and
To predict DAB more accurately, it is better to apply the Chapman-Enskog
temperature T,. become almost constant at the critical points of the compound or
theory than the above equations. For monatomic gases, which act ideally, we have
element, he went back to earlier literature and found empirical equations which
relate critical temperatures I;,, and volumes v;., to characteristics such as boiling
DAB
--
= 0.0018~83 T
P(<JAB) nn.AB
3 2
1 JMA
1 + 1 '
MB
(13.59) and melting temperatures, Tb and TM respectively. From these he developed several
equations for estimating s/Ks and a values for metal vapors. These have been used
to develop Table 13.7 for aAB values, and we may use the following equations for
where <JAB = !(<IA + aB) = collision diameter, A, (13.60)
s/K 8 :
nD,AB = collision integral for A-B mixture at dimensionless
temperature, T]iJ, for the Lennard-Jones potential. (13.61)
= 1.92 T M~K. s/K 8 (13.62)
* (KB)
TAB= - . T, Although Eqs. (13.61) and (13.62) give somewhat differing values of s/K8 , the errors
S AB
in DAB values, particularly at temperatures greater than TM are not great, and the
~ 10.0
estimates for DAB are valid within 20 % of the true value.
Example 13.2 Calculate the diffusion coefficient for iron vapor diffusing through
"'"'
II
~ 7.5
argon, at 1600C, assuming that pure iron is the source of the vapor.
Solution. Using the Chapman-Enskog equation, we obtain
0.001858(1873) 312 1 1
DFe-Ar = '1)(aFe-Ar)2QD,Fe-Ar 55.85 + 39.54.
We evaluate aFe-Ar from the data in Tables 1.1 and 13.7:
Fig. 13.18 Variation of collision integral with reduced temperature for the Lennard-Jones 25
E.T. Turkdogan, paper in Steelmaking, The Chipman Conference, J. F. Elliott (ed.), Addison-
(6-12) potential. Wesley, Reading, Massachusetts, 1962.
26
J. Hirschfelder, R. Curtiss, and R. B. Bird, The Molecular Theory of Gases, Wiley, New York,
24
R. B. Bird, W. E. Stewart, and E. N. Lightfoot, ibid. 1956.
-----e---r-------~--~-------
Tablel3.7 Estimated collision cross sections for metal vapors. binary gas systems and temperatures possible, is that of Fuller et al. 27 They
correlated 340 data points in a theoretically justified form, predicted DAB to
<Y,A within 10 % of the measured values 92.6 % of the time, and had an average error
Melting Boiling Molar volume, cm 3 /mol
of only 4.3 % and a standard deviation of 6.17 %. Their equation is
Metal point,
oc
point,
oc V(sol.) v(liq.) vb(liq.)
From
V(sol.)
From
vb(liq.)
D -
AB -
(1 x 10- 3 )Ti. 75
P(v113 + vjf3)2
J 1
MA
1
+ MB , cm 2 /sec, (13.63)
Ag 960.8 2163 11.02 11.60 13.06 2.72 2.74
AI 660 2057 10.49 11.36 13.26 2.67 2.76 where T = temperature, K, P = pressure, atm, MA, MB = molecular weight of
Bi 271 1477 20.76 24.07 3.37 species A and B, gm/gm-mole, and vA, vB = diffusion volumes, given for simple
Cd 321 765 13.34 14.01 14.89 2.90 2.87 molecules in Table 13.8. This approach is relatively simple and avoids the necessity
't Co 1493 2877 7.14 7.68 8.98 2.36 2.43 to evaluate the collision integral.
Cu 1083 2570 7.58 8.01 9.15 2.39 2.44
Fe 1535 2833 7.60 7.94 9.55 2.40 2.47
Ga 29.8 1983 11.43 Table 13.8 Diffusion volumes for simple molecules*
13.87 2.80
Hg -38.9 357 14.09 14.65 15.71 2.95 2.92
In 156.4 2087 15.94 16.33 19.98
Molecule v
3.07 3.16
K 63.7 760 46.0 47.10 59.36 4.37 4.54
Li 186 1317 13.27 13.40 16.76 2.89 H1 7.07
2.98
Mg 651 1103 14.66 15.47 20.30 3.00 3.18
D2 6.70
Na 97.9 883 24.04 24.79 31.07 3.52 3.66
He 2.88
Ni 1453 2816 7.11 7.57 9.02 N1 17.9
2.34 2.43
Pb 327.4 1717 18.35 19.39 . 22.47 3.22 3.29 02 16.6
Pu 637 3300 14.20 14.49 18.78
Air 20.1
2.95 3.11
Sb 630.5 1440 18.59 18.74 20.29
Ne 5.59
3.23 3.18
Sn 231.9 2770 16.51 17.04 21.63
Ar 16.1
3.11 3.25
Tl 303 1457 18.10 20.64
Kr 22.8
3.19
Zn 419.5 906 9.56 9.45 10.19 2.59 2.51
co 18.9
C0 2 26.9
* From E. T. Turkdogan, ibid. N 20 35.9
NH 3 14.9
H 20 12.7
In order to evaluate the collision integral, we need to calculate (c/Ks)Fe-Ar Cl 2 37.7
J
(!'Ks._) Fe-Ar = (!'Ks._) Fe (!'Ks._) Ar = )(3521)(124) = 655K. '
Br 2
S0 2
67.2
41.1
& Values of Din gas phases are usually in the range 0.1-10.0 cm 2 /sec. Figure
T* =(KB) T= 1873 = 2.86, 13.19 gives some representative data.
c Fe-Ar 655
and from Fig. 13.18, QD,Fe-Ar = 0.92, or 13.6 DIFFUSION THROUGH POROUS MEDIA
0.001858 (1873) 3 ' 2 As we have pointed out in previou~ chapters, metallurgists often encounter situa-
1 1 2
DFe-Ar = (1) (2.92)2(0.92) 55.85 + 39.54 = 4.00 cm /sec. tions where they must deal with the properties of solids containing some degree of
porosity. Gas diffusion through porous media is important in such fields as ore
Many other correlations have been developed for nonmetallic gases. The most
successful correlation, which makes the prediction of DAB over a wide range of 27 E. N. Fuller, P. D. Schettler, and J.C. Giddings, Indust. and Eng. Chem. 58, 19 (May 1966).
~--~-------------------- . ------
oo T, C distribution, and particle shape. It is not possible to predict r on the basis of material
000000 0 0 0
parameters, but reasonable estimates may be made by comparison with previous
' 10.0 -\O:::~~~~ ~ 0
N
-
0
0
results using similar materials, such as presented in Fig. 13.20.
u
1A '
' I\.'-
~ . ~ 1.0 o Glass spheres
~
o Sand (uncompacted)
'-..." ~......
~- 0.8
Carborundum
I":...'--.,:hit.
._ 0 Ci 0.7
T Sodium chloride
0 Soil crumbs
"' Pumice
Ee Talc
, '\. H20-02 ~........... 0.6
!:>. Kaolin
\.,.''-..CHr02 !j, , + Celite
0.5
1.0 '--CO o~~C0-02
L.....--22,,
"" '-
~" ,_
"' .... ""',.........._
."_~
-- -" _
H 20 H 2
0 2-H2-
co 2-H2-
~C02-H2-
=
0.4
Steel wool
V Vermiculite
v Mica
Sand (compacted)
........... Hg-H 2 -
0.3
'
I'..:::~~2 _/, Ar-N 2
,02-Air-
1
o~,~Cd-N 2
0.2
.
2 H 20-C0 2
~Hg-Air
0.1
5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0
,, t;.<;)
""
101r, K 1
Fig. 13.19 Diffusion coefficients in gas mixtures.
0.3 0.4 0.5 0.6 0. 7 0.8 1.0
reduction or roasting, vapor penetration into foundry sands, outgassing of powder Porosity. cm 3 / cm 3
metallurgy compacts, gas phase alloying of po~ders, and catalysis.
Fig. 13.20 Hydrogen-air diffusion in various unconsolidated porous media. (Adapted from
In general, gas phase diffusion through porous media occurs by one of two
L. N. Satterfield and T. K. Sherwood, The Role of Diffusion in Catalysis, Addison-Wesley,
mechanisms: ordinary diffusion or Knudsen diffusion. (A third mechanism, surface Reading, Massachusetts, 1963.)
diffusion, may be important at low temperatures, but for most elevated temperature
processes, it is negligible.) The size of the pores through which diffusion is taking
place determines whether ordinary or Knudsen diffusion predominates. With When the gas density is low, or the pore size small, then the molecules have a
ordinary diffusion, the pores are large relative to the mean free path of the gas higher probability of collision with the pore walls than with each other, similarly
molecules. Knudsen diffusion is important when the pores are small in relation to as in the flow of molecules in a vacuum system. An analysis of the flow of molecules
the mean free path. through a cylindrical pore of radius r under a concentration gradient yields
When ordinary diffusion prevails, the total diffusion path, from one plane in
the porous media to a parallel plane, is longer than when no aggregate is present,
.
lA
2 -(dCA)
= 3r~ dx .
because of the irregularity of the porosity. Even after correcting for the decrease
in crosscsectional area available for diffusion, using the void fraction, w, we must ~ By using Eq. (1.4) for ~' the average speed of the molecules, and comparing it with
make a further decrease in the DAB value, in order to yield the effective interdif- Fick's first law, we obtain the Knudsen diffusion coefficient:
fusivity. We accomplish this by introducing a new factor, the tortuosity, so that
DAB,eff =
DABW
-r- (13.64)
Dx = 9700r ;z, cm 2 /sec, (13.65)
l
---~---------e-r""-...-----~-----~-------
longer predominates and Knudsen diffusion becomes important. In the first test, Table 13.9 Knudsen diffusion and flow in porous media*
we may estimate the pore radius and compare it to the mean free path of the gas A.,
given by Material Technique Gases T,K r,A T
_I
(3.4)(0.45)
(4.0)
= 0.383 cm 2 /sec,
14
or approximately one-tenth the value in the gas phase alone.
DIFFUSION IN SOLIDS
PROBLEMS
13.1 Show that the units in Eq. (13.1) are as indicated. Do the same for Eqs. (13.2) and (13.3).
," ,~ 13.2 Discuss the reasons why self-diffusion data must apply to homogeneous materials only.
' 13.3 Read one of the references of Footnote 1 and derive Eq. (13.5).
13.4 Look up the article by Compaan and Haven (Table 13.1) and summarize the method This chapter, in which the primary aim is to obtain solutions to Fick's laws of
used to derive the correlation coefficients in Table 13.1. diffusion, is almost identical to Chapter 9 where we presented solutions to heat
13.5 Work out the units in Eq. (13.20). conduction problems. Therefore, much of the groundwork laid in Chapter 9 will
13.6 Derive Eq. (13.37), after reading Zener's article. reappear in the following discussions. We shall begin by presenting some classical
13.7 Using the method of Fuller, Schettler, and Giddings, estimate the diffusion coefficient approaches used for determining diffusion coefficients in solids, and then consider
for a COr02 gas mixture at 1-atm pressure and 700K and compare the result to the data in some applied problems involving diffusion in solids as the rate-limiting step.
Fig. 13.20.
13.8 Assuming a tortuosity of 2.0 and a void fraction of 0.25, calculate the effective diffusion 14.1 STEADY-STATE DIFFUSION EXPERIMENTS
coefficient for CO-C0 2 in a reduced iron oxide pellet at 800K.
As an example of the application of Fick's first law, consider an experiment in
13.9 PbS has an NaCl-type structure. Would you expect the self-diffusion coefficient of Pb or
S to be higher? How would you expect the addition of Ag 2 S to affect the diffusion coefficient which an iron tube is held in the isothermal part of a furnace. A carburizing gas is
of Pb, knowing that the defects in PbS are predominantly Frenkel defects on the Pb sublattice, passed through the inside of the tube, and a carburizing gas of a different com-
and that the undoped PbS is an n-type semiconductor? How would Bi 2 S3 additions affect it? position is passed over the outside. Steady state is reached when the carbon
[Ref.: G. Simovich and J.B. Wagner, Jr., J. Chem. Phys. 38, 1368 (1963).] concentration at each point in the tube wall no longer changes with time. By this
13.10 Describe the conditions under which the following terms are applicable: time, the appropriate differential equation for steady-state diffusion through a
(1) self-diffusion, \ cylinder can be derived from shell balances (Problem 14.7). If the diffusion co-
(2) tracer diffusion, efficient of carbon in iron is a constant, independent of composition, then
(3) chemical diffusion,
(4) interstitial diffusion,
(5) substitutional diffusion,
~ !!_
r dr
(r dC) =
dr
0. (14.1)
(6) interdiffusion coefficient, and
The solution to this equation is given by its heat transfer analog from Eq. (7.65):
(7) intrinsic diffusion.
C - C2 ln(r/r 2 )
(14.2)
ln(r i/r 2)'
where r 1 and r2 are the inside and outside radii of the tube, and C 1 and C2 are the
corresponding concentrations of carbon at these surfaces. Thus a plot of C versus
In r should be a straight line. However, for carbon diffusing in y-iron, the slope of
such a plot, as shown in Fig. 14.1, becomes smaller on passing from the low-carbon
side to the high-carbon side. Therefore, the diffusion coefficient must be a function
of composition, Eqs. (14.1) and (14.2) do not apply, and we must approach the
problem somewhat differently.
In addition to the fact that oCjot = 0, steady state also means that the quantity
473
p 0.12 l.6 ~ -!of---
0
0 ~
0
0.10 1.4
""' p, P2
0.08
1.2 "
0
u
t::
Glass
'. c,
Metal foil
l.O .e
0.06 0.8 u"
Fig. 14.2 Experiment for diffusion of hydrogen through metal foils.
0.6
0.04 The foils are very thin, and so it is extremely difficult to determine the concentration
0.4 as a function of distance through the foil. The experimental results therefore consist
0.02 of a measured steady-state flux, the hydrogen pressure drop across the foil, and the
0.2
foil's thickness. To obtain D from the data, we take the value of C in the metal at
0 0
0.24 0.26 0.28 0.30 0.32 0.34 0.36 each gas-metal interface as the solubility S that would exist in equilibrium with
-logr the gas.* From Sievert's law, we know that for equilibrium between gas and the
metal
Fig. 14.1 Steady-state carbon concentration profile through a hollow cylinder of iron at
1000C. (From R. P. Smith, Acta Met. 1, 578 (1953).)
(14.7a)
and
of carbon passing through the tube per unit time is constant and independent of
S2 = Kpy2, (14.7b)
r. Thus in which K includes the equilibrium constant for the reaction
J = 2nrlj,, (14.3) H 2 (g) = 2H(in solution), (14.8)
where I = length of the cylinder, j, = local flux, and J = quantity of carbon and p 1 and p 2 are the partial pressures of hydrogen on both sides of the foil with a
passing through the tube wall per unit time. Since thickness b, as shown in Fig. 14.2.
The gradient dC / dx can then be expressed in terms of the pressures:
dC
j, = -Ddr, (14.4)
dC S1 - S2 K c c
dx = b =-g-(vP1 -vP2). (14.9)
we may express Eq. (14.3) as
Combining Eqs. (14.6) and (14.9), we obtain the flux:
r[-DdCJ = _!_,
dr 2nl -DK C C
jx = -b-(v P1 - v P2). (14.10)
or
dC -J In relation to the diffusion of gases through solids, the term permeability, P,
(14.5) is often used defined by
d In r 2nlD
For a given experiment, we can measure J and I, and if the carbon concentrations P = DS = DKj"P, (14.11)
within the tube wall are determined by chemical analyses, then we can determine so that
D from the slope of the plot of C versus In r. (14.12)
Similar experimenrs have also been performed for determining the diffusion
coefficient of gases through metals such as, for example, the diffusion of hydrogen * This is true under the conditions whoo. the solution of gas in the surface of the metal occurs
through a metal foil in Fig. 14.2. According to Fick's first law, the flux of hydrogen much more rapidly that the rate at which the diffusing species leaves the surface and enters the
through the metal is bulk metal. Experimentally, we check this assumption by determining the fluxes for two
thicknesses of foil under the same pressure drop and temperature. If equilibrium does exist
dC at the interface, then ~C is the same for both cases, and the flux is inversely proportional to the
jx = -D- (14.6)
dx thickness.
..
~---.,-.-., - - - - . , --------
I"
Permeabilities have also been measured for gases diffusing through other From Table 14.l and Eq. (14.16), P* = 8.4 x 10- 6 cm 3 (STP)/sec-cm-atm 112 .
materials. Table 14.2 contains some of these data. Therefore
I
_l_
In more general terms, where r 1 is not specified, but l is still 100 cm, quantity {3 of the tracer is plated as fl thin film Ax' thick on one end of a long rod of
2
tracer-free material. The rod is then annealed at the diffusion temperature of
J = -1.755 x 10- interest Since D* is a self-diffusion coefficient and does not depend on position,
log (r ifr 2 ) for such an application, Fick's second law is
Both results are shown below: ac =
- D*-
a2 c (14.18)
or- 2
ox
{) 1.0 1.5
! 1.0 Suppose we take a second tracer-free rod and butt-weld it to the plated end
il:' (without any diffusion occurring), and then carry out the diffusion anneal. Accord-
f-< 0.9
~ ing to Eq. (9.89), we see that the solution is
0.8
-~) 1:u
.....; 0.7
C(x, t) = C;
en;: (-x2)A,
exp D* ux. (1419)
.
0.6 2....; nD*t 4 t
0.5
Here C; is theconcentration of the tracer in the plated material whose thickness
0.4 is Ax'.
0.3 Since C; Ax' is the quantity of tracer material plated as the thin film, we write
0.2 the solution
0.1 .
C(x, t) =
{3 (
JrJ5*t
exp -Dx* '
2) (14.20)
0.0 2 nD*t 4 t
5
r 2 , cm which describes the spreading by diffusion of a thin plate source into an infinite
sink. This is illustrated in Fig. 14.3.
14.2 TRANSIENT DIFFUSION EXPERIMENTS
~ l.25
Under most circumstances it is not possible to carry out steady-state experiments (..)
I
----------------
---~
If diffusion is now allowed to take place, the material that would normally diffuse in
f
the negative x-direction is reflected at the x = 0 plane, and moves in the positive c I
r---weld
x-direction. The concentration at any x in the + x domain is then given by the c,-----.--1 I interface
superposition of the original solution for x > 0 and the reflected solution for Initial
x < 0, or distribution
C(x, t) =
f3
fiiJ*i ( x2)
exp 4 D*t (14.22) x=O
Distance x
(a)
N 1.0
cS y
. ~ 14.2.2 Diffusion couple with constant jj
I
I -
\..) \..) 0.8
The previous case has only a limited practical value because the initial distribution
usually is not an infinitely thin source, but rather it occupies a finite region. This 0.6
case is exemplified in Fig. 14.4a by butt-welding two bars of A-B alloy of con-
centrations C1 and C2 . If 15 is independent of composition, and the bars extend 0.4
far enough in the positive and negative domains to be cons,idered infinite, then
Eq. (14.18) applies with the boundary conditions: 0.2
Specifically at x = 0.05 cm, C = 0.35 % C. The results for various locations as well or finally we can get
as for the longer times are:
A. dC
- 2 dA. =
d (-dC)
dA. D dA. .
(14.27)
Consider the diffusion couple depicted in Fig. 14.5a. Then for C(A.), we recognize
that
for A. = - oo, (14.28a)
0.3
x, cm I= 0
~ ,~
and
c,
oc oA. (de} 1(de} (14.26b)
OX = OX dA. =t dA. .
1 12
0 x
Substituting into Eq. (14.25), we obtain (c)
Fig. 14.5 (a) Initial conditions. (b) Definition of location of Matano interface after diffusion
has taken place for a time t. (c) The integral and the slope obtained in order to calculate /5
at composition C.
and
----f Problem 14.6. Furthermore, Jost 1 has pointed out that it is not necessary for a
single phase to exist over the entire range of the diffusion region. A discontinuity
for A.=+ oo. (14.28b) in C(x) and D(C) exists where an intermediate phase is formed. The only condition
We then solve Eq. (14.27) by integrating between C = C1 and C = C: required when applying the Boltzmann-Matano technique is that the concentra-
tions in both phases on either side of the interface between the phases are indepen-
c c dent of time.
--1
2
J A.dC = D
-dC- I
dA. (14.29)
Example 14.3 A diffusion couple made of pure Nb welded to pure U is held at
c, c,
800C for 49 days. The resulting concentration profile, shown below, is obtained
Since the concentration gradient goes to zero as C approaches C1 , the right-hand with a microprobe analyzer. Note that an intermediate phase bis formed between
side of Eq. (14.29) is simply D(dC/dA.). Then the two solid solutions, y 1 and Yz. Calculate i5 at 90 at.% U.
!?- 100
- 1 Jc (14.30)
I
I I
D=-(dC) A.dC. i:i 90 I I
2 - c, .2 r-1 I
dA. ' 80 Matano iI i1Marker interface
From Eq. (14.29), we obtain the definition of the Matano interface; since the con-
""
~
0
70
interface
I
I
I
I
u I
- centration gradient also goes to zero as C approaches C 2 , Eq. (14.29) gives us the :i 60 I I
I
I I
additional condition that 50 'Yi 'Y1
40
(14.31) 30
\
20
Since experimental data are available only at some constant time t, Eqs. (14.30) 10
and (14.31) can be written in terms of x and t, and the relationships used for
o'-t1~-'-7
calculating i5 from the measured concentration profile are 2-'-o!:---'-:';-2-'--4-7-'--6~~s';--'-~1~0--'-':-"17
4 2-'-;-'14c;--'-7167-'-~1~s~2'0
xX 10 3 in.
c
- 1 1 J From N. L. Peterson and R. F. Ogilvie, Trans. AIME 218, 439 (1960).
JxdC = 0. (14.33)
This is found by trial and error, and pl<:i-ced on. the figure as shown.
At 90 % U, dC/dx is evaluated:
c,
dCjdx = 1810 at. %/cm.
In Fig. 14.5(b), the plane x = 0 is given by the line that makes the two hatched areas Then
equal. We calculate the value of i5 at a given C by measuring the cross-hatched
area f~1 x dC and the reciprocal slope at that point, dx/dC. The diffusion coefficient cf 9 x dC = -0.132 cm_ at.%, by graphical means;
15, found by applying Eq. (14.32) in this manner, is the interdiffusion coefficient C= 100
discussed in Section 13.2.3.
In addition, if we place inert markers at the original plane of welding, we can 1 W. Jost, Diffusion in Solids, Liquids, and Gases, Academic Press, New York, 1960, page 76,
also determine the intrinsic diffusion coefficients. This is left for the reader to do in and also W. Jost, Z. Physik 127, 163 (1950).
--~---------------"-----""
49 days 3600 sec 24 hr where (:} is (C - Cs), and all the boundary conditions can be written in a homo-
t = = 4.24 x 106 sec. geneous form. According to Eqs. (9.78) and (9.79), we see that
hr day
Thus X = c 1 cos AX + c 2 sin AX, (14.36)
- -(-0.132) / and
D (90% U) = 2(4.24 x 10 6 )(1.810 x 10 3 )
G =exp (-A 2 Dt). (14.37)
= 8.6 x 10- 12 cm 2 /sec.
Boundary condition (14.35b) requires that c 2 = 0, and then when we apply (14.35c)
c 1 cos AL = 0 results. This is satisfied by A = (2n + l)n/2L where n is any integer
14.3 FINITE SYSTEM SOLUTIONS from 0 to oo. Hence
~" ~) '
The solutions to Fick's laws presented thus far represent useful cases in many
~
2
ll [-(2n + 1)2n Dt] [(2n + l)n x] (14.38)
situations, but there are also other problems of interest. This is the case when the u = n~O An exp 4 L2 cos 2 L '
system boundaries are close together, relatively speaking, and the sink cannot be
considered infinite or semi-infinite, as the effect of the diffusion process on the where the An's are now the constants involved. The initial condition, O(x, 0) =
composition is felt at the furthest point in the material prior to the end of the O; = C; - C., remains to be satisfied and when substituted into Eq. (14.38), it yields:
diffusion treatment. This condition can arise quite often when small parts are
exposed to a gaseous environment, and there is diffusion of the gas species into the 00
(2n + l)n x
part. Conversely, metal parts must often be degassed. When dealing with these ei = I A cos - (14.39)
n= l,odd n 2 L
situations, it is usually safe to assume that the rate-limiting step of the overall mass
transfer is the diffusion of the gas species into the solid. Hence, we seek solutions of If/we apply Fourier's analysis to Eq. (14.39) as we demonstrated previously for
Fick's second law with a surface concentration imposed at time zero and main- Eqs. (9.48), (9.59), and (9.83), we obtain
tained constant.
To illustrate this case, consider the diffusion into or out of a slab of infinite (-It 4 e (14.40)
length and thickness 2L. Initially the slab has a uniform concentration C;, and
An = (2n + 1); i
then its surfaces are raised or lowered to Cs and maintained constant. Hence, we Thus the solution we seek is
are seeking a solution to Fick's second law with constant D (or simply D):
ac oC 2 !!_ = C - Cs =~ I (- lt exp [-(2n + 1) n 2 2
Dt] cos (2n + l)n ~- (14.41)
- = D -2
ot ox (14.34) 0; C; - cs n n=O 2n + 1 4 L2 2 L
The initial and boundary conditions of interest are Equation (14.41) is useful for describing concentration profiles as a function of
time. However, the total amount of material that diffuses into or out of the slab
C(x,O) = C;, (14.35a) is often of more interest, particularly in experimental work where this is the only
measurable quantity. So, the average concentration C is required:
ac (14.35b)
dx (0, t) = 0,
and
-
C = L1 JL Cdx. (14.42)
0
C(L, t) = Cs. (14.35c)
Carrying out this operation, using ~q. (14.41) for C, we obtain the relative change
The solution to this problem can be determined in the same manner as the in average composition for diffusion into a slab:
solution of the heat conduction equation in Section 9.4. By separation of variables
8 ~
2 2
C - Cs exp [-(2n + 1) n -Dt]
it has the form : 1
--- - - 2 L... (14.43)
(:} = X(x) G(t), C; - Cs - n n=o(2n + 1) 2 4 L2
. -.. . . . . . . ----
---,~-
-----------~~;r-~------~~
This expression is good for diffusion into or out of a slab. If we take the first term Table 14.3 The relative change in average composition for the basic shapes
in the series /
Diffusion in a slab of semithickness, L
c- c 8
C. _ ~ = n 2 exp (-t/r), (14.44) LC.: C(x, 0) = C;,
' s
B.C.: C(L, t) = C.,
where r is the time constant for the diffusion process (r = 4L 2 /n 2 D), it is apparent
that a graph of log 8 versus (t/r) is a straight line and that we can obtain D from the ac
-(0, t) = 0.
slope. We plot Eq. (14.43) in Fig. 14.6, along with diffusion into or out of cylinders ax
and spheres (see Table 14.3). For long times (Dt/L 2 > 0.05), the first term of the Solution
series is sufficient, and a straight line relationship is obeyed.
Cs _ ~ ;,
2 2
E- 1 [-(2n + 1) n Dt]
r.:;lr.:;
1.0
C - z L.. (2 1)2 exp 4 Lz . (14.43)
I I 0.8 Ci - s 7t n=O n +
1<:..l G 0.6
Diffusion in solid circular cylinder of radius, R
0.4
0.3
LC.: C(r,O) = C;,
0.2
/ B.C.: C(R, t) = C.,
0.10
0.08
ac
-(0, t) = 0.
0.06 or
0.04
Solution
0.03
0.02 E- cs 00
I 4 exp --~-,
-2 (-eDt) (14.45)
C; - Cs n=!~n R.
0.010
0.008 where~" = 2.405, 5.520, 8.654, 11.792, 14.931, when n = 1, 2, 3, 4, 5, etc.*
0.006
Diffusion in spheres of radius, R
0.004
0.003 The same set of initial and boundary conditions as for the cylinder above:
0.002
E-
---=2
Cs 6
I
00
1
-zexp
(-n nDt)
2
2
2
(14.46)
0.001 0 C; - Cs n n= 1 n R
0.1 0.2 Q.3 0.4 0.5 0.6 0.7
/j)f Dt *~.are roots of the equation J 0 (x) = 0, where J 0 (x) is the Bessel function of zero order.
L2 orR2
Fig. 14.6 The relative change in average composition for the basic shapes. R = radius, and after 40 hours of vacuum outgassing treatment at a temperature where DH =
L = semithickness. 1.0 x 10- 5 cm 2 /sec, assuming an initially uniform distribution.
For diffusion into or out of simple multidimensional shapes other than the
Solution
infinite plate, infinite cylinder, or sphere, we handle the problem in the same manner a) Consider this to be an infinite plate. Then
as we treated heat transfer to or from these shapes in Section 9.5. We can combine 5
Dt (1 x 10- )(40 x 3600) = 0.056.
product solutions to yield the solution for the shape of interest.
L1 (2 :x 2.54) 2
Example 14.4 Calculate the fraction of hydrogen remaining in
a) a 4-in. thick slab of steel, 10 ft long x 4 ft wide, From Fig. 14.6, we have
b) a 4-in 2 square billet of steel, 10 ft long, and c - cs= 0.74.
c) a 4-in 2 square billet of steel, 8 in. long, C; - Cs
--------.a:r111U~IUll~UllI:l"U'II~-p~~;.,'t;;;:,-n1n1-a-11IUr111e-11I1.1;:"TIU~~ ...: : r I - - - - - -
b) The desired solution can be obtained as the product of the infinite plate boundary is at x = M, and at this boundary Cfi and Cf represent the equilibrium
solutions for the two 4-in. dimensions: concentrations that coexist in phases II and I, respectively, at the temperature
C-C under consideration. In both phases, diffusion takes place, and there is exchange
c. - ~ = (0.74)(0.74) = 0.55. of mass from one phase to the other at the interface.
' s In both phases, Fick's second law applies. Hence
c) In this case, first evaluate Dt/L 2 for the 8-in. dimension:
ac, -_ DI 8ax2C1,
Tr
2
x > M, (14.47a)
Dt (1 x 10- 5 )(40 x 3600)
L1 (4 x 2.54)2 = 0.028.
and
From Fig. 14.6, we get
(C;c -- c.)
x <M. (14.47b)
= 0.80.
c. s"ctim.
D1 and Du are the diffusion coefficients of the diffusing substance; both are assumed
Therefore \ to be independent of composition within each phase.
The next condition assumed to hold in this situation where diffusion is the
C-
- -
8 c '
= (0.80)(0.74)(0.74) = 0.431. rate-controlling step, is that the concentrations on either side of the interface are
C; - c. related by an equilibrium expression of the form
C[=Kq, (14.48)
14.4 DIFFUSION-CONTROLLED PROCESSES WITH A MOVING INTERFACE
where K is the partition ratio between the' phases. (Later, in Chapter 16, we shall
Many metallurgical reactions and processes involve diffusion steps in conjunction consider the situation where equilibrium is not reached at the interface, that is,
with reactions at phase boundaries. One result of these transient processes is the where diffusion does not completely control the overall rate.)
motion of the boundary between the phases. The analysis of the diffusion process
Finally, we shall always conserve mass as transfer across the interface from
is somewhat complicated due to the presence of more than one phase. First, we
one phase to another occurs. The net difference in the flux of the diffusing solute
shall discuss some of the general aspects of the problem, including the boundary
entering and leaving the interface equals the solute added to phase II by virtue of
condition at the moving interface, and then present applications in carburizing and
the moving interface, that is,
decarburizing.
14.4.1 General aspects (ac,)
D, -
8
- Du -
8X
(acu) =(Cu* - C,)-d
* dM (14.49)
X x=M x=M t
In the general situation, two phases are in contact as in Fig. 14. 7. The moving phase
14.4.2 Formation of a second phase on the surface of an initially homogeneous
c first phase
One application of the preceding basic assumption is the decarburization of y-Fe
with the formation of IX-Fe on the surface in the temperature range 723-910C, or
the c<trburization of IX-Fe to form y-Fe in the same range, as originally presented by
Wagner. 2
II
We presume thatFick's second law holds in each phase, and assume equilibrium
conditions at the 1X-y interface. For reference, Fig. 14.8 gives a portion of the
iron-carbon phase diagram. Consider the situation depicted by Fig. 14.9(a).
Initially, the alloy contains C;-carbon and is heated to the austenite region.
x 2
C. Wagner, unpublished notes from Course 3.63, M.I.T., Cambridge, Massachusetts, 1955.
Fig. 14.7 The concentration profile of two phases coexisting during transient diffusion. See also W. Jost, ibid.
l_
oc = M, C11 = C~ (constant), it
B1 and B 2 are constants of integration, and since at x
follows that the argument of the error function must be constant' at x = M.
910 Hence
M = 2f3J"D;;i, (14.53)
where f3 is a dimensionless parameter t_o be determined. Substituting Eqs. (14.51)-
(14.53) into the mass balance equation (14.49), we get
723
q = c. - B2 erf (/3), (14.54)
and
C~ 1 O.Q25 C* 0.83
(14.55)
,'Y
wt%C where
Fig. 14.8 Portion of the iron-carbon phase diagram.
and
B -pz B -pzq,
C* - C* - _ze__ - 1 e . (14.56)
C;
II I - Jn/3 Jnpfl
"' Eliminating B1 and B 2 between Eqs. (14.54)-(14.56), the parameter f3 is fo4nd by
trial and error as that value which satisfies
c, C;
(II) (I)
(C* _ C*) = (C. - Crt) _ (ct - CJ . (14.57)
0 M
(a)
x
0 M
(b)
x 11 1
Jnf3eli2
erf f3 Jnp J<f>efi 4> erfc (/3 2 112
)
Fig. 14.9 Concentration profiles for cases where a second phase II forms on the surface of an 1.0
initial homogeneous phase I.
~"
" 0.0
b'
"
L>"
Depending on the carbon potential of the atmosphere, a surface composition of -1.0
c. is established. Therefore the conditions to be satisfied, in addition to Eqs. (14.48)
-2.0
and (14.49), are
LC.: C(x,O) = C;, (14.50a) -3.0
and -4.0
C 1 = C;
.
+ B1 (1 - erf ~)
2y1D 1t
x>M, (14.51) -7.0
-8.0
and -9.0
-1.0 0.0 1.0 2.0 3.0
0 < x < M. (14.52) a
2
Fig. 14.10 Graph of t:5 values as a function of Jnt:5e" erfc <5.
-- ----- - - - - - - - - - - - - " - - - -- ----------- - ~---- - - ----------""--..... ---r""'--n-,.., ---"'T"J1J1"EJ"..._~~.-a"~"'T7;;::J
The denominator of the center term contains the function of fJ, fJeP erf fl, which we
2
The correct value of fJ has been bracketed. Further trials eventually result in
have already encountered in Chapter 10, and we may evaluate it by means of fJ = 0.144. Now we can calculate the layer thickness M:
2
Fig. 10.6. The right-hand term contains a function of the form Jn(Je" erfc (J (in
this case fJ</> 112 = (J), and we can evaluate it from Fig. 14.10. Solution of the
M = 2fJ~ = 2(0.144)(2 x 10- 6 x 1.8 x 103 ) 1 12 = 0.0173 cm.
practical problem of predicting the rate of advance of the interface would require
14.4.3 Formation of a single-phase layer from an initial two-phase mixture
specifying the compositions and diffusion coefficients, and determining the value
that satisfies Eq. (14.57). Then we simply relate Mand t by Eq. (14.53). We can This situation is best exemplified by the carburization or decarburization of a steel
also determine B1 and B 2 using Eqs. (14.54) and (14.55), respectively, and then the that is initially a two-phase mixture; four possible cases are shown in Fig. 14.11.
concentration profile, as given by Eqs. (14.51) and (14.52), is completely specified. The interface M is now where the concentration of carbon in the surface phase
reaches the concentration required for equilibrium with the second phase. In the
Example 14.5 A thin shell of 0.40 %C alloy steel is austenitized at 800C in an
two-phase region, the average composition C0 is assumed to be uniform, which
. i atmosphere of CO and C0 2 that is in equilibrium with 0.01 % ~- Under these I requires, in effect, that the grain size is small and that second-phase dispersion is
conditions some a-Fe forms on the surface. After 30 min, how thick will the layer
uniform.
ofa-Fe be?
Solution. In this case: C; = 0.40 %, Cs = 0.01 %, q = 0.24 %, Cjlj = 0.02 %,
D11 = 2 x 10- 6 cm 2 /sec, D1 = 3 x 10- 8, cm 2 /sec, and</>-= 66.6.
I
I
723 < T< 910C 723 < T< 9!0C
1 c,
Cl\ ----Co
Co
'( '
(II) ' + "Y
(U) '( + '
0 M M
x x
0:02 0.24 0.4
T>723C
%C T< 723 C
----Co
Eq. (14.53) gives the thickness of the a-Fe layer once fJ has been evaluated by
Co
means ofEq. (14.57):
(0.01 - 0.02) (0.24 - 0.40)
(0.02 - 0.24) = f(fJ) - f(fJ<jJlf2) , c, (II) ' + Fe 3 C 7+ Fe 3 C
0 M 0 M
0.01 0.16 x x
0.22 = f(fJ) - f((J).
Fig. 14.11 Examples of the formation of a single-phase layer on a surface when the initial
Trial and error is used, along with Figs. 14.10 and 10.6. Let fJ = 0.2 ;f (fJ) = 0.045 material is a two-phase mixture with average composition C0 .
from Fig. 10.6. Then, (J = /3 112 = (0.2)(8.1) = 1.62, and f ((J) = 0.85 from Fig.
14.10. Substituting these into Eq. (14.57), we have
Making the same assumptions as before, that is, D's are constant and interface
reactions are so fast that only diffusion controls the rate of movement of M, we set
O.OlO - 016 = 0 034 < 0 22
0.045 0.85 . . . down the conditions that describe the situation. Equation (14.47b) is obeyed in
the region 0 < x < M, with the boundary condition that Cu(O, t) = C5
Try fJ = 0.1; f(fJ) = 0.011, (J = 1.215, f((J) = 0. 70. Again we give the solution by Eq. (14.52). The equilibrium at M is simply
described by
O.OlO - 0.1 6 = 0.68 > 0.22.
0.011 0.70 Cu(M, t) = C~, (14.58)
----- --------- ----------------.---
and the material balance at the interface takes the form concentration in equilibrium with the liquid. A cast structure is exemplified in
Fig. 14.12 showing the repetitive pattern of microsegregation. In the case at hand
0C11 ) dM
-Dn ( ~ = (q - Co)-d (14.59) we presume the alloy to be a single phase.
uX x=M t
Equation (14.53) describes the locus of M with time. Then substituting Eqs. (14.53)
and (14.52) into (14.58) and (14.59), we obtain
Cs ~, C~ = B 2 erf /3, (14.60)
and
* - Co
Cn = -B2- exp ( - 132 ). (1~.61)
, Jnp First to
soli~ify
tare given, or we can calculate D if M and tare measured. In turn, we can determine
Fig. 14.12 A dendritic structure showing the coring or microsegregation of the alloying element.
B 2 from Eq. (14.60). These relationships have been tested against the data of
L is one-half of the dendrite arm spacing.
Bramley and Beebe, 3 with satisfactory agreement.
Example 14.6 Calculate the depth of decarburi~ation of an 0.20 % C, Cr-Mo steel, During homogenization, the alloy naturally tends toward a uniform concentra-
after 1 year of exposure to severe decarburizing conditions at 950F. At 950F the tion (Fig. 14.13). We need only examine what happens within one dendritic element
diffusion coefficient of carbon in this steel (ferrite) is 1.0 x 10- 9 cm 2 /sec, 4 and the since the profile is repetitive, and there is no net flow of solute from any dendritic
region to the next.
steel is a two-phase mixture (a + Fe 3 C).
Solution. One year equals 3.02 x 10 7 sec. Cs~ 0.00%C, C 0 = 0.20%C, and
c~ = 0.02%.
As-cast
- 1- (O.OO - 0.0 2) = f3eP' erf f3 = 0.0626. .! distribution
M = (2 x 0.23)(3.02 x 10- 2 ) 1 12 Fig. 14.13 Dendritic element, 0 < x < L, used for describing homogenization kinetics.
= 0.078 cm. Tq, describe the homogenization kinetics, a solution to Fick's second law is
needed that satisfies
14.5 HOMOGENIZATION OF ALLOYS J.C.: C(x, 0) = f(x), (14.63a)
During solidification of alloys, coring occurs, because the rate of diffusion for most
B.C.:
ac (14.63b)
alloying elements in the solid state is too slow to maintain a solid of uniform ax (0, t) = 0,
3 A. Bramley and G. H. Beebe, Carnegie Scholarship Memoirs 15, 71 (1926). ac t > 0. (14.63c)
4
OX (L, t) = 0,
R.R. Arnold, M.S. Thesis, Univ. of Wisconsin, 1968.
l
-- ------------------------
2
An = L f
L
nnx
f(x) cos L dx. (14.65)
ingots the dendrite arm spacing could typically be 300 which would have to be
held at 2150F for about 20 hours to reduce [J to 0.2 for chromium. In large com-
mercial ingots, dendrite arm spacings are much larger and homogenization is even
0
more difficult to achieve .
." i In Eq. (14.64), C0 is the overall or average alloy content, and we evaluate the
An's by Eq. (14.65) using f(x) as the initial solute distribution. 2. Significant homogenization of substitutional elements is possible at reasonabl~
A useful parameter to describe the homogenization kinetics is the residual temperatures and times only ifthe material has fine dendrite arm spacings ( < 50 )
segregation index [J which is defined as which result from rapid solidification.
(14.66) 3. Interstitial elements (such as carbon in steel) diffuse very rapidly at austenizing
temperatures.
where CM= maximum concentration, that is, CM= C(O, t); C~ = initial maximum Homogenization studies have also been carried out for 4340 low-alloy steel 6
concentration, that is, C~ = C(O, O); Cm= minimum concentration, that is, Cm= and 7075 aluminum alloy. 7 In Reference 7, the analysis and discussion of homo-
C(L, t); and C~ = initial minimum concentration, that is, C~ = C(L, 0). For no genization emphasizes the dissolution of a nonequilibrium second phase during
homogenization, [J = 1, and after complete homogenization [J = 0. solutionizing.
We can find CM - Cm by applying Eq. (14.64) to x = 0 and x = L, and
performing the indicated subtraction. The residual segregation index can then be Example 14.7 An ingot of 52100 steel goes through the processing schedule
written as indicated below. Estimate the residual segregation index of chromium for the
00 material in the center and the outside surface of the finished bar. Neglect the small
2 - L An exp - n1n2 Dt)
( L2 amount of diffusion that occurs during blooming, rolling, and final cooling. Near
[J = n-1,3 ... odd
(14.67)
cZt - c~
">< 1.0
"'
"O
= 0.8
.,,"s
Oil
0.6 ""
"'8, 0.4
M 0
'~"" 0.2
I
-----------
the surface of the original cast ingot, the dendrite arm spacing is 40 , and in the
center, it is 800 . The diffusion coefficient of chromium in this steel at these
1 ~
u
NE
u
s-
28
24
I
\
x
temperatures is given by Q
20 \
\
D = 2.35 x 10- 5 exp [ - l 7,300/T(K)J. \ I
I
16 \ ,.__Center
\ I
\ I
12
8
urface
4
2 2 0
i.--<--W--22 0 10 20 30
Ingot
soaking Time, hr
During proc.essing, assume that the dendrite spacing decreases in proportion
hr
_ _ Surface to the changes of linear dimensions. Based on this, L(t) is given in the table below.
_ _ _ Center
as
D/L 2 , 8
sec- 1
>
6 ll
We can then use this value of Dt/L 2 in Eq. (14.67). For a continuous nonisothermal
situation, we compute the total magnitude of Dt/L 2 to be used as 4
t
Dt
Lz = 1
Lz
JD(t) dt. 2
0
0
For hot-working in which D and L are both time dependent, Dt/L 2 should be
Time, hr
evaluated as
Under area I 112 x 10- 5 hr sec- 1
t
Dt Underareali 141x10- 5 hrsec- 1
Lz f
0
D(t) dt
[L(t)] 2 .
Dt
Total 253 x 10- 5 hrsec- 1
L 2 (surface)= 253 x 10-s x 3600 = 9.10.
First determine D(t) given the thermal process schedule and D(T).
!
According to Fig. 14.14, this value of Dt/L 2 indicates that the surface material through the oxide layer so thatj0 2- = 0. Thus the oxide thickens because the metal
would be homogeneous since b ~ 0. cations diffuse through the oxide so that the oxidation reaction can proceed. At
In the center of the ingot, since the dendrite spacing is 20 times the spacing of any instant, the flux of cations through the oxide is given by Fick's law (Eq. 13.2):
the surface, then (neglecting small differences in the thermal history)
Dt Dt 1
. .
] = (JA2+) = -D (ac)
ax
L 2 (center) ~ L 2 (surface) x
202
Here Dis the diffusivity of the cation, and C is the cation concentration. If the oxide
9.10 layer is thin, then we approximate the concentration profile of the cation as linear,
= 400 = 0.0228. and can integrate Fick's law as
M CM
,' i From Fig.14.14, we see that b ~ 0.27, and a significant amount ofmicrosegregation
Temains in the material. j Jdx = - J D dC.
0 Co
If we consider the situation in Fig. 14.15, we can derive an expression for the or
rate of increase in the oxide thickness M. In this example, the metal is divalent, and
2
the oxygen anions, having a large ionic radius, diffuse only at a negligible rate M = 2k't, (14.71)
_l
-:iu'I____ --umuSJon-in-sonos----------- --- - - ------------- - - - - - - - - - - - - - - - - - - - - - - - - - - - ...------pf';U'--------------- ronnanon-orsurrace-raTDISn-1ayers-:iu:.
where k' is a constant with units cm 2 /sec, known as the Tammann scaling constant 8 or in terms of the free energy for an ideal solution
or the Pilling and Bedworth constant. 9 The parabolic nature of the rate of change
of the oxide thickness with time is apparent (14.76)
Experimentally, it is usually more convenient to measure weight gain rather
than the oxide thickness. Then
where; is the chemical potential (per atom) of species i. If the compound formed
Am has the stoichiometric composition A 0 Bb where A is the cation, then the scale
A= Mpo, (14.72) consists of ions according to the dissociation reaction:
(14.73) (14.78)
where kP is the practical parabolic scaling constant, (g 0 2 ) 2 /cm 4 -sec. Usually, we ( In order to obtain an expression for the total flux in terms that we can measure,
obtain experimental data by weight gain measurements, and take straight-line we must replace the expressions involving A+, B-, and e. We first consider the
behavior when we plot (Am/A) 2 versus t as an indication of diffusion-controlled equilibria
oxidation.
(14.79a)
Such data alone, however, do not indicate what species is responsible for the
major material flow, that is, whether the metal is diffusing out from the oxide- and
metal interface or the oxygen is diffusing in. Wagner 10 has extended this simple
(14.79b)
expression to express kP in terms of diffusion coefficients of the migrating species.
Combining Eqs. (13.30), (13.47), and (13.49), we obtain the relation At equilibrium
t-a (14.80a)
n-B- = -'-, (14.74)
' ' (z;e)2 and
where n; = concentration of the migrating species, B; = mobility of the species, B = B- - zBe. (14.80b)
a = total electrical conductivity of the compound, t; = transference number of
For the compound in general, we get
species i, z; = valence, and e = electronic charge. Now, using Eq. (13.42) and
substituting Eqs. (13.47) and (14.74), we obtain the flux ft;:
(lll.81)
(14.75)
from the Gibbs-Duhem relationship. Then, eliminating A+, B-, and e from
Eqs. (14.78) and (14.76) and using Eqs. (14.80a, b) and (14.81), we obtain expressions
for the particle fluxes:
8
G. Tammann, Z. anorg. u. allgem. Chem. 111, 78 (1920).
9
N. B. Pilling and R. E. Bedworth, J. Inst. Metals 29, 529 (1923). ftA + = zBtAtea (oB)' A ions/cm 2 -sec, (14.82)
10
C. Wagner, Atom Movements, Amer. Soc. for Metals, Cleveland, Ohio, 1951, page 153. zAe 2 z~ ox
J_
and ~---,-:~~""'o~ cannoto~ua1"'~""-1~~~~:=~~~~=-
<<A + i.l
because it would cause a charge imbalance. In an electronic semiconductor, the
product te(tA + tB) is a very small number, and k, is small. In this case, where
B ions/cm 2 -sec. (14.83)
te = 1.0, if we assume that mobilities BA and BB are equal to BJ and B;, respectively,
Since the growth rate of the compound is the sum of the particle fluxes (although we can substitute Eq. (13.47) into Eq. (14.89) and obtain
either may go to zero), we obtain
k, = ~
e N zB0
Ja(tA + tB)te dB, (14.86) 02
~ P02 =I atm
where k and i are the chemical potentials of Bat the inside and outside (metal-
Solution. Since D5 is much less than D~b we simplify Eq. (14.90) to
gas) interfaces of the oxide, respectively; k, has the units of g-equivalents cm- 1-
sec-1 and is related to the parabolic rate constant kP by
2
_ 2p 0 (at. wt B) k
kp - r (14.87)
r(mol. wt AaBb)
The Tammann scaling constant is related to k, by The term nBr/N0 has the units of equivalents/cm 3 . In this case, we have
k' = 2(at. wt B). k 2pNiO 2(7.44) . 3
(14.88) = = 0.199 eqmvalents/cm .
PobzB r mo.1 wtNiO 74.69
If we assume that the anion exists as B2 in the gaseous state, and there is ideal Alternatively, if we take zB as electronic charges/B atom, and nB is the concentration
gas behavior, we can substitute of B atoms/cm 3 , then we must divide by Faraday's constant, 96,487 couls/equivalent,
and multipiy by the charge on an electron, 1.602 x 10- 19 couls/charge. In this
case, we get
into Eq. (14.86) and obtain
(14.89)
We note that although te may approach 1.0, and (tA + tB) may approach zero, the - (6.00 x 10 )(2)(1.602 x 10- 19 )
-
22
2 x 96,48 7
Jvg,2DN).3
*
d log Poi.
pi
*Note that numerically r = z 8 , but that their units differ. 02
___ _L
---rq:o,-----..----p,;u ---------- - - - - - - - . --run11a:nu11-01-~una\...~-u1n11311-ia:y~
Thus either way Tablel4.4 Parabolic oxidation constants for various metals
f
0
Po,
0.199 2 Conditions
k, = - - (D~J2.3 d log Poi. Metal Oxide kP, (g 0 2) /cm 4 -sec
2 Temp., C p02 , 1 atm
., '1''?117~1
Ni/NiO
I
Log P02 atm where Vzr is the molar volume o.( Zr and Vzroi is the molar volume of Zr0 2 The
diffusing species in the oxide is th~oxygen anion moving in from the gas phase to
equal to 16 x 10- u. Therefore the Zr0 2 -Zr interface. At the interface, some of the oxygen dissolves in the metal,
( 0 ~99 )(2.3)(16
and some reacts to form more Zr0 2 Beyond the interface and into the metallic
k, = x 10- 11 ) = 3.66 x 10- 11 equivalents/cm-sec. phase, for x > M, we have
I
------- -.
----------------~-----~------...-~-~------~
where xis the distance from the original interface, and x' = x - M, where x' is the constant (independent of concentration) and equals 2 x 10- 10 cm 2 /sec at 1000C. Markers
distance from the oxide-metal interface. Thus are inserted at the original interface and move along during the diffusion process at a com-
position of 20.205 atom fraction Zn. Determine the intrinsic diffusion coefficients of copper
x' -
erfc [ - - + -Vz,- - jf'] and zinc at 20.205 atom fraction Zn.
14.2 The term "banding" is used to describe chemical heterogeneity in rolled steels that shows
C = C 2JD;;i: Vz,02 Do . (14.95) up as closely spaced light and dark bands in the microstructure of steel. These bands represent
e erfc [ -Vz,
- -
Vz,o 2
jf'J
Do
areas of segregation of alloying elements during freezing of the ingot. During rolling the
segregated areas are elongated and compressed into narrow bands. Assume that the alloy
concentration varies sinusoidally with distance after rolling according to the sketch below.
I 14.4 One side of an iron sheet, 0.01 cm thick, is subjected to a carburizing atmosphere at
1700F such that a surface concentration of 1.2 % carbon is maintained. The opposite face is
11
maintained at 0.1 % carbon. At steady state, determine the flux (g-moles/cm 2 sec) of carbon
N. L. Peterson and R. E. Ogilvie, Trans. AIME 218, 439 (1960). through the sheet
12
D.R. Gabe, JISI 204, 95 (1966).
13
A. U. Seybolt, Trans. ASM 29, 937 (1941). a) if the diffusion coefficient is assumed to be independent of concentration (D = 2 x
14
J. S. Kirkaldy, Can. J. Phys. 36, 899 (1958). 10- 7 cm 2 /sec);
15
J. S. Kirkaldy, Can. J. Phys. 37, 30 (1959). b) if the diffusion coefficient varies as shown in the graph on page 512.
-7
5 x 10 zero atom fraction carbon. At steady state, make a plot of the composition profile in the sample
indicating clearly compositions and respective distances.
4
The original thickness is 1 mm and density changes during the experiment may be neglected.
At 800C, it is known that the diffusion coefficient of carbon in iron is given by:
3 D = 10- 6 cm 2 /sec in ferrite (a),
D = 10-s cm 2 /sec in austenite (y).
2
.," ~)
1.2
% Carbon 'Y +graphite
800
14.5 A composite foil made of metal A bonded to metal B, each 0.01 cm thick, is subjected to
t atm of pure hydrogen on metal A's face; the ot~er side, metal B's face, is subjected to a perfect 0.15 738 C
vacuum. At the temperature of interest and 1 atm of hydrogen, the solubility of hydrogen in
metal A is 4 x 10- 4 g per cm 3 of A and in Bit is 1 x 10- 4 g per cm 3 of B. It is also known that
hydrogen diffuses four times faster in A than B and that A and B do not diffuse in each other. 700 ex+ graphite
At steady state draw the concentration profile of hydrogen across the composite foil.
14.6 A gold-nickel diffusion couple of limiting compos_itions XNi = 0.0974 and XNi = 0.4978
is heated at 925C for 2.07 x 10 6 sec. Layers 0.003 in. thick and parallel to the original
0 2 4
interface are machined off and analyzed.
a) Using the data tabulated below, calculate the diffusion coefficient at 20, 30, and 40 at.% Atom fraction carbon
nickel.
14.9 In order to make a transformer steel with the proper hysteresis loop, a low silicon steel
b) Suppose that markers are inserted at the original interface and move along during the
sheet is to be exposed on both sides to an atmosphere of SiC1 4 which dissociates to Si(g) and
diffusion process at a composition of 30 atom fraction nickel. From this, determine the
Cl 2 (g). The Si(g) dissolves in the steel up to 3 % at equilibrium.
intrinsic coefficients of gold and nickel at 30 atom fraction nickel.
a) Indicate what equation and what boundary and initial conditions would apply in order to
Slice No. a/o Ni Slice No. a/o Ni Slice No. a/o Ni Slice No. a/o Ni calculate how long you would have to hold the sample before the composition is essentially
uniform at 3 % across the sample.
II 49.78 22 33.17 29 21.38 39 12.55 b) Using the data in Fig. 13.12, calculate the time required to reach 95 %ofthe 3 %Siat 1800F.
12 49.59 23 31.40 30 20.51 41 11.41
14 47.45 24 29.74 31 19.12 43 10.48
16 44.49 26 25.87 32 17.92 45 9.99
18 40.58 27 24.11 33 16.86 47 9.74
19 38.01 28 22.49 35 15.49
20 37.01 37 13.90
21 35.10 38 13.26
14.7 By making use of the shell balance technique, derive an expression for diffusion through a
hollow cylinder.
14.8 A thin sheet of iron at 800C is subjected to different gaseous atmospheres on both of its
surfaces such that the.composition of one face is at 4 atom fraction carbon and the other is at
When we combine Eq. (15.3) with Eq. (15.1), we obtain a form ofFick's first law for
a binary solution:
15 (15.4)
MASS TRANSFER IN FLUID SYSTEMS 15.1 DIFFUSION THROUGH A STAGNANT GAS FILM
Consider the system shown in Fig. 15.1 where liquid A is evaporating into gas B,
and a constant liquid level at x = 0 is maintained. At the liquid-gas phase inter-
face (x = 0), the gas phase concentration of A is that corresponding to the vapor
pressure of A at that temperature.* For simplicity, also assume that the solubility
of B in liquid A is negligible and that the entire system is maintained at a constant
In Chapters 2 and 7, we demonstrated the development of differential equations temrerature and pressure. At the top of the tube, a stream of A-B gas flows past
pertinent to momentum and energy transport in simple fluid systems. In this
slowly)hereby maintaining a constant concentration of A at x = l which is less
chapter, we shall consider how to formulate and describe elementary diffusion
than the liquid-gas interface concentration. Therefore, a concentration difference
and mass transfer problems in fluid systems. We use practically the same procedure
of A exists between x = 0 and x = l, which causes diffusion. When the system
in this situation as we did previously; we develop a differential equation, and a
attains a steady state, there is a net motion of A away from the evaporating surface
solution containing arbitrary constants evolves. These constants are evaluated by and the vapor B is stationary.
applying boundary conditions that specify the concentration or the mass flux
Under these conditions, despite the fact that gas Bis stationary, there is bulk
at the bounding surfaces. Again, we demonstrate the principles involved by
motion of fluid since A itself is moving, and its motion contributes to the average
considering specific examples, but first let us reconsider the general situation, as
velocity. Thus we refer to Eq. (15.4) for the flux of A with NBx = 0. Solving for
outlined in Section 13.2.2.
NAx we obtain
Species A ilf-a gas stream moving in the x-direction is under the influence of a
concentration gradient, also in the x-direction. The molar flux of A relative to N - - CDA dXA.
Ax - 1 - XA dx (15.5)
stationary coordinates is then made up of two parts: CAv: which is the molar
flux of A resulting from the bulk motion, and jAx = - CDA(oXA/ox) which is the Gas stream of A and B
diffusive contribution. Thus 1.0 XB
I
xiB /i
I
I x'A
I
(15.1) I
I
I
I
Here v;' is the local molar average velocity in the x-dir~ction, and vAx is the velocity I
I
I
B A x
v: = ~( CAvAx + CBvBx) * This, of course, implies that equilibrium is maintained at the interface, i.e., from a very
simplified mechanistic viewpoint, atoms (or molecules) can readily leave the liquid state and
enter the gas phase. This assumption is valid except at very high diffusion rates where the
= ~( NAx + NBx). (15.3) rate of transfer of atoms across the liquid-gas interface is not able to keep pace with the
exhaustion of the atoms away from the interface.
514
------------------ ----,,.__.-- - - - - - - - - ---- ---
A mass balance on a unit volume .1x of column height (see Fig. 15.1) for steady The information that is most often sought after is the rate of mass transfer at
state is the liquid-gas interface. We obtain this by using Eq. (15.5):
(15.6)
CDA (dX~)
in which S is the cross-sectional area. In the expected manner, we divide by ~x
NAxlx=O =
_
0
1 - XA dx x=O
= CDA In (1 - X,\_ ),
l 1 - XA0
(15.14)
and take the limit as .1x ~ 0:
or
dNAx = 0
dx
(15.7) N
Axlx=O
= CDA (X1 -
(XB)In l
X,\_) (15.15)
Substitution of Eq. (15.6) fo~ NAx yields
where (X B)10 is the log mean of the terminal values of X B.
!_( CDA . dXA) = O.
dx 1 - XA dx
(15.8) X B1 -XB0
(X) - (15.16)
B In - In (X1/X~)
As pointed out in Chapter 13, diffusion coefficients for isothermal gas solutions
are very nearly independent of concentration; also, C is constant for an ideal gas Equation (15.15) is somewhat more appealing because a characteristic concentra-
mixture at constant temperature and pressure. Hence we simplify the derivative tion difference X1 - X,\_ over a distance l is evident. For a gas in which species A
further: is dilute, Eq. (15.15) reduces to
(15.9) (15.17)
Two successive integrations can be made directly, resulting in which could have resulted from originally ignoring bulk motion and expressing
(15.10) the flux of A simply as
Figure 15.2 illustrates one technique which We use to determine the vapor pressure where~nd r2 are the roots of
of a metal (liquid or solid). Argon, as a carrier gas, passes over the sample which
is at the temperature corresponding to the vapor pressure being determined.
2 v:
r - DA r = 0. (15.23)
This gas, containing the saturation concehtration of the metal vapor, enters the
exit tube at z = 0, and at the cool end of the exit tube the metal condenses out and Thus, r 1 = 0 and r2 = v:/DA> and the solution is
deposits where we can collect it for subsequent mass determination.
We evaluate the arbitrary constants by using Eqs. (15.24) and (15.21a and b):
xo
(15.25)
dNAz = O (15.19)
dz The concentration -profile can then be written:
We may choose either Eq. (15.1) or Eq. (15.4) to represent NAz; here we select
Eq. (15.1) from which we write exp v*l) - exp (v*z)
(t_ ~
(15.27)
(15.20) v*l) 1
exp (~A -
We can certainly consider that the argon-metal gas solution is ideal, and if the
temperature variation between z = 0 and z = l is small, then C and DA are con- We evaluate the flux at which the- metal vapor enters the exiting gas stream at
stants, and Eq. (15.20) takes the form z = 0.
d 2 XA
--
v:
dXA
---=0 (15.21)
dz 2 DA dz
, . - - - - - - - - - .::i-,,;u-IV1ass"lTansrer-in'Tlmo-sySiems- -- -- --"--------- - - - - - - - - - - 1 " : ' . ) ; - 4 : - - , . . , , _ _ , , . - - - - - I - : J ; : J - - - - - - - - - - - - u1rrusnm-1ncu-a-rau1ng-i111u10-nnn-::i:n
quantities by
P1 _ sNAzlz=Orexp
0 _
1
- ] (~) (15.29)
P - XA - SCvi (vii) , '
exp -
DA
where P~ is the vapor pressure of A, and Pis the total pressure. Preferably, the effect
of diffusion should be negligible for best experimental results due to the uncertainty
of the diffusion coefficient and because we would have to assume an experimental 10 20 30 40
set up that corresponds to the mathematical formulation. /,cm
The real value of the analysis lies in the group vii/DA which indicates how to
set up the experiment, so that the effects of diffusion may be ignored. Ifvil/DA 2:: 5, At temperatures where p~n is greater than negligible values, and n> nAn
the effect of diffusion may be ignored because the last term in Eq. (15.29) is 2:: 0.993 the curves should be shifted even further to the right for better results.
and < 1.0. To insure sufficiently high values of vii/DA, the experimentalist should
provide a small diameter tube between z = 0 and z = l, and use argon or nitrogen
as the carrier gas, rather than hydrogen or helium, since D in the lighter gases is
15.3 DIFFUSION INTO A FALLING LIQUID FILM
larger than in the heavier gases.
In this section, we shall consider a fluid system moving in such a way that the
Example 15.2 An experimental apparatus is being constructed to study the velocity distribution is unaffected by diffusion into the fluid. Figure 15.3 shows a
thermodynamics of Mn-Cu alloys by measuring the Mn vapor pressure over the film of liquid B falling in laminar fl.ow down a wall. Gas A is absorbed at the
molten alloys at 1400K. In order to use the transport technique, what exit tube liquid-gas interface; we shall restrict the situation to that where the penetration
dimensions and argon gas fl.ow rates should be used? distance of A into Bis small relatively to the film thickness. We wish to calculate
the amount of gas absorbed after the film travels a distance L.
Solution. The criterion that provides the most direct experimental measurement First, we develop a mass balance on component A. The gas-liquid interface
of P~0 (the equilibrium pressure over the alloy) is concentration of A at all points along the film, is the saturation value C1 ; thus A
diffuses into the liquid which initially contains less than the saturation amount
pMO = pNMn, of A. As the film drops, the liquid is exposed to C1 for a longer time and more
n N penetration of A into the film results. We see, therefore, that CA changes both
where NMn is the number of moles of Mn condensed out and N = NMn + NAr
x
with and z, and we select the unit volume: Ax by Az by unity in they-direction.
Then the mass balance for A is simply
= total moles of gas passing through the exit tube over some period of time.
This is essentially true when v*l/DMn-Ar 2:: 5.0. (15.30)
--r=~w_h_e_n __d_e_a_l-in_g_w-it_h_e_n-er-:n,port '" a c~~=~~~=~~~=~
to obtain the velocity profile vz(x); similarly, we need to describe the velocity for
the analogous situation of mass transfer in a convective system. For a falling film,
we have already worked this out in Chapter 2 in the absence of mass transfer at the
L
fluid surface, and we know that the results for fully developed flow are
v(x)
where [J is the film thickness, and x is the distance into the film from the gas-liquid
surface.
If, as indicated in Fig. 15.3, A has penetrated only a slight distance into the film,
then for the most part species A sees only vmax Further, since it does not penetrate
very far, we can consider that the liquid is semi-infinite. These conditions would
Fig. 15.3 Absorption into a falling film; hold, for example, for short contact times. With these approximations we write
the differential equation and the boundary conditions:
Dividing by AxAz, and performing the usual limiting process, we get
ocA 0 2cA
oNAz + oNAx = 0. (15.31) Vmax oz ~DA OX2 . (15.36)
oz ox .
We now have to insert into this equation the expressions for NAz and NAx For Atz = 0, CA= C~, x ;:::-: 0, (15.37a)
the molar flux in the z-direction, we write
at x = 0, cA = c1, L;:==-: z ;:::-: 0, (15.37b)
(15.32)
at x = co, CA= C~, L;:==-: z ;:::-: 0. (15.37c)
and neglect the diffusive contribution, realizing that A moves in the z-direction
primarily due to bulk flow. In addition, for small increases in the concentration Note that we may alternatively view z/vmax as the time t, over which a moving
of A, v: ~ vz. Finally, we give NAz by the simplified expression slice of liquid has been subjected to the surface concentration C1. Thus, we
recognize the solution to Eq. (15.36), subject to Eqs. (15.37a), (15.37b), and (15.37c),
NAz = CAvz. (15.33) as the golution for the temperature distribution in a semi-finite solid, initially at a
For the molar flux in the x-direction, we write uniform temperature, which is suddenly subjected to a new constant surface
/ temperature. Then, referring to Eq. (9.110), we have
ocA //
NAx =-DA~+ cA. (15.34)
ux //
//
/ (15.38)
That is, in the x-direction, A is transported primarily by diffusion, there being
almost no bulk flow in the x-direction due to the small solubility of A in B. Sub-
Now knowing the concentration profile, we proceed to determine the local
stitution of Eqs. (15.32) and (15.34) into Eq. (15.31) yields the differential equation
mass flux at the surface, x = 0:
-~diffusion
for CA(x, z):
(15.35) (15.39)
I
L
- ----~-....---~--1::.;'1--------- - - ---------
the average rate of A transferred per unit across the entire surface between z = o transfer, a mass-transfer coefficient for transfer of A into or out of a phase is
and z = L being defined in terms of the diffusive contribution normal to the interface:
f
L 0
k _ }A DA(oCA/ox)x=o
- (15.45)
NAxlx=O = L1 NAxlx=O dz M - c1 - CAco c1 - CAco
0
Here, the superscript 0 refers to quantities evaluated at the interface, and CA 00 to
= 2(C1 _ C~)J.DAvmax. (15.40)
some concentration of A within the fluid, usually the bulk concentration. Note that,
nL while kM in Eq. (15.45) is defined in terms of the diffusion flux at the surface, in
general, at interfaces involving a fluid phase, there is the additional contribution to
Pigford 1 solved the complet~ equation mass transfer caused by bulk flow. We define the mass-transfer coefficient here only
in terms of the diffusive contribution, rather than of the total flux N1. This is
(15.41) because the coefficient so defined is somewhat more fundamental, since we might
expect the diffusion flux to be approximately proportional to a characteristic
and obtained the result: concentration difference as indicated by Eq. (15.45), whereas the bulk flow contri-
bution can be relatively independent of any concentration difference. Similarly,
-L 0 when both heat and mass transfer occur, it is advantageous to retain the definition
c~ - c~ = 0.7857e-5.1213~ + 0.1001e-39.318q + ... (15.42) of the heat-transfer coefficient given by Eq. (8.1), which considers only the conduc-
CA - CA
tion flux.
where 1/ = DAL/b 2 vmax' and c;; = bulk average composition of the liquid at L. In the limit oflow mass-transfer rates, as is often the case, we may neglect the
For small values of rJ, corresponding to short contact times or very thick distortion of the velocity and concentration profiles by mass transfer, and the
films, we obtain .bulk flow term is negligible. Then
(15.43) i k _
M -
N1
c1 - CAco
(15.46)'
and for long times, we have This equation is definitional only, and we must evaluate it by means of various
~; =~i ~ 0.7857e-' ""' l\1-f (15.44) v * analytical expressions for the flux. As the first example of applying Eq. (15.46),
consider the results of diffusion into a falling film in Section 15.3. We evaluate
the local mass-transfer coefficient relating the rate of mass transfer of A into the
liquid when the time of contact is short, by substituting Eq. (15.39) into Eq. (15.46).
15.4 THE MASS-TRANSFER COEFFICIENT
_ NAxlx=O
kM.z- _- JDAVmax (15.47)
In Chapter 7, we analyzed simple problems of heat transfer with laminar convec- O ___ ,
CA - CAco nz
tion, and formulated the temperature distribution from which we calculated
heat-transfer rates by evaluating the heat fluxes at the fluid-solid boundary. With o'r in terms 0f dimensionless groups, and recalling that Vmax = ~V,
the fluxes, we wrote expressions for the heat-transfer coefficients, and we saw that
Nu = f (Re, Pr), and gained insight into what was to follow in Chapter 8 where we = [3 /Vz fv .
vv:i
kM.zZ
(15.48)
presented the correlations for heat transfer in turbulent convective systems. DA v~v--;
Having considered diffusion in the presence of forced convection in Section
15.3, it is convenient to introduce the mass-transfer coefficient. As we have men-
The group kM,zz/DA is called the S_herwood number, Sh, or alternatively the mass
transfer Nusselt number, NuM. The Reynolds number should be easily recognized.
tioned, we may treat the movement of a species as the sum of a diffusional contribu-
The Sch~t number, Sc, which is defined by
tion and a bulk flow contribution (see Eq. 15.1). To be analogous with heat
v
1
R. L. Pigford, Ph.D. Thesis, University of Illinois, 1941.
Sc=-, (15.49)
D
-----------------
is the analog of the Prandtl number encountered in heat transfer. Most available Substituting Eq. (15.44), we obtain
forced-convection mass-transfer correlations are in the form
. ' kM
vc5
= L [In (e 5 1213 q - In 0.7857)]
Sh =/(Re, Sc, geometry),* (15.50)
as, for example, in the situation above. Specifically, we could write Eq. (15.48) as
=T
vc5 [5.121311 + 0.241]
Shz = ~ Re:' Sc 2 1 2
' . (15.51)
(15.55)
In this case, by subscripting with z, we emphasize that local values are being
considered. If we used Eq. (15.40) instead of Eq. (15.39), we would define an By rearranging this expression, we get
average mass-transfer coefficient over the film length L.
kMb
(C1 - C~)(4D~~max) 1/2 D
A
=Sh~ 3.42, (15.56)
kM =---~-~--- which is similar to the results given in Table 7.1 where the Nusselt number was
(C1 - C~)
(15.52) found to be a constant for fully developed laminar flow. We consider that this
= f6DJ'. equation is applicable at Re ( = r /17) ~ 25, where r is the mass flow rate per unit
y---;I: width of film.
We can then write this as Having been tested for absorption of gases into liquids flowing down wetted-
wall columns, Eq. (15.56) has been found to somewhat underestimate the actual
mass-transfer coefficient. At low Reynolds numbers, this is now understood to be
partly due to the so-called Marangoni effect, in which upward-directed surface
or in dimensionless form tension forces counteract the downward-directed gravitational forces, causing
rippling and turbulence on the surface and an increase in transfer which is not
(15.53) anticipated in the simplified development described above.
where the subscript L indicates that the quantities are averaged over the entire film Example 15.3 A method for degassing molten metals involves exposing a thin film
length. of metal to vacuum by allowing it to flow continuously over an inclined plate.
In the case of long contact times, where Eq. (15.44) applies, the rate at which A Calculate the average hydrogen concentration of a ferrous alloy with an initial
is absorbed in the distance dz is concentration of 1 cm 3 H 2 {SIP) per cm 3 of alloy flowing down a plate 100 cm
long and 15 cm wide, which is inclined at an angle of 1 deg from the horizontal.
VbdCA = kM(C1 - CA) dz. The concentration of hydrogen at the surface exposed to the vacuum may be taken
Over the entire length of the film, the absorption rate of A is as zero. The desired film thickness is 1 mm. Data are as follows: p = 8.32 g/cm 3 , ~
11 = 6 cP, Da = 1.3 x 10- 4 cm 2 /sec. ~
Solutio'Pz. Using Eq. (2.14), we find the average velocity:
v~~
/v e 'YL
The integration yields
V _ (8.32 g/cm 3 )(980 cm/sec 2 ) (0.1 cm)2(cos 89) _
- ( _2 I ) - 7.90 cm/sec.
3 6 x 10 g cm-sec
(15.54) The contact time is v~ ~cl
L lOOcm
* The product ReSc which we often encounter in the literature on mass transfer, is sometimes
called the Peclet number, Pe.
-6 .:::: v= 7.90 cm/sec
= 12.7 sec .
(3.14)(100) 1.8
= 1.71x10- 3 cm/sec.
,\ J\!/I 0
~- '1~
I tU
1!()~ 1.6 e _ .p
,, 1.4
-~
Then
*~ Mass transfer
3 3 3
\ \ \ jH 2 = (1.71 x 10- cm/sec)(l cm H 2 /cm alloy)= 1.71 x 10- cm H 2 /cm -sec.
~"
. Total content removed per cm 2 of exposed surface is = jH 2 (contact time)
~
3 3 2
---
into given phase
~~~0- cm 3 H 2 /
2
= of given phase
~
2 0.6
Cv cm film.
e
Initial total content = (1cm:1f 2 /cm 3 alloy) (0.1 cmtalloy/cm 2 film) 0.4
1+
c~ - c~ _ (NA + NB) (15.59)
depicts the unit volume to which we apply the integral mass balance.
N - exp k The amount of A flowing intoihe element is
A --
--- COA M
I
NA+ NB
I (15.67)
I
I
I
w,,tl i
l I 0
I I
: WA,x+flx
5c ----,.-- Equation (15.67) is an exact analog of Eq. (7.36). If we assume a concentration
I
I profile analogous to Eq. (7.37), that is,
I
I
6x CA - C1 3( y) _~(I_) 3
(15.68)
%
x x+6x
CAoo - c1 = 2 J: 2 Jc '
Fig. 15.5 Integrated convective contributions to the concentration boundary layer over a which satisfies the conditions
flat plate. B.C.1: at y = 0, cA = c1, (15.69a)
B.C.2: CA= CAOO' (15.69b)
and the amount leaving is then by substituting the assumed velocity distribution, Eq. (2.105), and the con-
I
centration distribution, Eq. (15.68), into Eq. (15.67), and following the procedure
-D (acA)
A ay y=o 3 DA
The mass transfer into the unit element across the phase boundary at y = 0 is kM,x = (15.71)
c1 - CAoo 2 Jc
.
}Ayly=O L\x = -DA (acA)
-a L\x. (15.64) Thus, we combine Eqs. (15.70), (2.107), and (15.71) to obtain
y y=O
Shx = 0.323V&JRe":. (15.72)
We now formulate the mass balance for component A:
These solutions are valid for most fluids, including liquid metals, because
jAyly=O L\x + WA,x + WA,l = WA,x+.ix (15.65)
(Jc/J) 1 for metals. Table 15.1 gives typical magnitudes of Schmidt numbers for
Substituting Eqs. (15.61)-(15.64) into Eq. (15.65) yields fluids.
Table 15.1 Typical magnitudes of Prandtl numbers
and Schmidt numbers
(15.66)
Pr Sc
The integrals in Eq. (15.66) are split, that is,
Gases 0.6-1.0 0.1-2.0
Liquids 1-10 102-103
f =J+f
0 0 de ' Liquid metals 10-2 10 3
-----~-------------------
The average mass-transfer coefficient for transfer from a flat plate to a fluid 1
is given by =-(WAX+ WBJ
p
L
kM = _!_
L f kM,x dx = 0.646 DA 1 3 1 2
L Sc ' Re L ' ' (15. 73)
, Now we can proceed to develop a mass balance for the volume element. The
various contributions to the mass balance of component A are
0
Accumulation of mass of A in the volume element
or
(15.78)
15.6 GENERAL EQUATION OF DIFFUSION WITH CONVECTION
Finally, by combining Eqs. (15.77) and (15.78), we develop the diffusion equation for
In this section, we summarize the general approach to the law of mass conservation component A:
in the volume element L\x Ay Az, depicted in Fig. 2.4, through which a fluid con-
taining A in solution is flowing. In the following expressions, pA is the mass -opA +~'-PAV= v
..., *
..., p DAVPA (15.79)
ot
concentration (for example, g of A/cm 3 of total solution) as defined by Eq. (13.1);
WAx is the total mass flux of A in the x-direction and is composed of a diffusive term As is usually the case, simplifications are utilized more frequently than general
and a convective term. Specifically equations. Often, one can assume constant mass density and DA> and make some
simplification. For constant p and DA, Eq. (15.79) becomes
WAx = -pDA ( OPl)
OX + PAVx = PAVAx (15.75)
OpA 2
Tt + vVpA =DAV PA, (15.80)
where p is the density of the entire solution, Pl is the mass fraction of A, and
vx is the local mass average velocity in the x-direction; thus, we define vx suc:h that or if divided by MA (molecular weight of A), we get
the total mass flux of all components in the x-direction is made up of the sum of the
n component fluxes in the same direction:
n
-oCA
ot
+ vVCA =DAV CA.
2
(15.81)
PVx = L
i= 1
P;V;x (15.76)
The left-hand side of this equatioq is DCA/Dt, showing direct similarity with Eq.
For a binary, in order to illustrate, we write (7.90) which is the basis for the numerous analogies between heat and mass trans-
port in fluids with constant p.
The above analysis could have been made equally well in terms of molar
' fluxes such as we have used previously.
--"
I
Equation of continuity for component A: For constant C and DA, Eq. (15.83) takes the form
Diffusion equation for A in solution: This equation is usually applied to low-density gases at constant temperature and
pressure. The left-hand side of this eq_uation cannot be written as DCA/Dt because
of the appearance of v* rather than of v. A more simplified form of the above
equations, which is used for diffusion in solids or stationary liquids (v = 0 in
Eq. 15.81), or for equimolar counterdiffusion in gases (v* = 0 in Eq. 15.84), is
Table 15.2 The equation of continuity of A in various coordinate systems Fick's second law of diffusion:
Momentum,
ovx ovx o2vx (15.87)
Cylindrical coordinates: Vx OX + Vy oy = v oy2 '
acA+ (v -+v
acA 0 --+v
I acA acA) oT oT o2T
-at r ar r ae z-az Energy, vx ox + vy oy = rx oy2, (15.88)
L
A typical set of boundary conditions that may be specified is: 0
~ _,,,::,
16
1.0 ::,I o
I _,,, ~
tTv'
I
c] c]
I
II
-e.
Pr--+ Sc.
hi~
I 11
h h' Then, the local Sherwood number is
0.4
"" - Pr = 0 7 Sc = 0 7
Shx = 0.332 Sc 0 343 Re; 12 , (15.91)
0.2 -- Pr = I : ' Sc = I .
and the average Sherwood number
6 8 IO 12
ShL = 0.664 Sc 0 343 Ref 2 . (15.92)
y
x v'Re.: 15.8 CORRELATIONS OF MASS-TRANSFER COEFFICIENTS
FOR TURBULENT FLOW
Fig. 15.6 Temperature and concentration profiles in a laminar boundary layer on.a flat plate
for Pr = Sc = 0.7, and Pr = Sc = 1. Curves for Pr = Sc = 1 also represent velocity profiles. We have seen in the previous sedions that many forced-convection mass-transfer
(From J.P. Hartnett and E. R. G. Eckert, Trans. ASME 79, 247 (1957).) situations are completely analogous to heat-transfer situations, and the appropriate
he~t-transfer solutions apply with simple changes of notation, namely:
These profiles show a dependence on the mass flux (v 0 ~/V00 ). Mass
transfer away from the plate (positive v0 ) gives flatter profiles as would be true
if the solid surface were porous and a gas diffused upward through the plate, or if a T--+ CA,
liquid passed through the porous plate and evaporated. On the other hand, mass Pr--+ Sc,
transfer towards the plate (negative v0 ) gives steeper profiles; this situation can
be obtained if condensation occurs at the surface, or suction is applied to a porous Nu--+ Sh.
L
~n addi~o~,we may ""ume that the mulb foe naturnl convection '"ult~g fro~---:hi
d~s;~f ~u%~ne~:s caused by mass transfer may be correlated by a mass-transfer
celationhip h~~~n
found to adequately demibe the rate of depoition of
3
metallic solutes from liquid to solid, and the rate of dissolution of carbon into
iron melts. 4
In flow around curved surfaces, such as spheres and cylinders, f /2 greatly
GrM = ge(xA - XAoc,)L 3/v 3, exceeds jH and jM. However, the analogy still holds between heat and mass
where eis the concentration coefficient of volumetric expansion defined as: transfer, so that jH and jM should be equivalent. To illustrate this, consider the
heat-transfer correlation given in Chapter 8 for forced convection around a sphere
e-- P~( axA
op )
r.
of radius R:
2 3 (8.10)
2hR/kJ = 2.0 + 0.60 Re}f Pr}1 .
~n this case, we may use correlations for heat transfer to yield mass-t ~ d
if the substitution rans er ata Translated into the jH form it becomes
fl ow m round tubes:
Cont~nuing the analogy, we define a mass-transfer j-factor jM for fully developed 0.011:-----+-----=="""-:1--""'"-.?~~..::c--~--r-:.,.------I
' '
. Sh f
lM = --(Sc) 213 = - (15.93)
Re Sc 2
Ifflo~ is n?t full~ developed, we use Fig. 8.2 where we take L/D into account and
~ubstituteJM for lH Epstein 2 used Eq. (15.93) to compute the corrosion rate ~fan 0.0001 101 10 5 106
uon tubde by m?lten mercury; apparently, the mass transfer of iron into the mercury Re=DVooP
stream etermmes the rate of this process. 7)
In the case _of flow past a flat plate, from the results of Chapter 8, we write for Fig. 15.8 Comparison of mass-, heat-, and momentum-transfer to spheres. (1) f /2; (2) Chilton-
average values m laminar flow: Colburn factorjH; (3)jM for cinnamic acid-water system; (4)jM for 2-naphthol-water system;
. . f 0.664 (5) jM for uranium dissolving in cadmium. (Data from E. D. Taylor, L. Burris, and C. J.
}H =JM = - = --, (15.94) Geankoplis, I. & E. C. Fundamentals 4, 119 (1965); and T. R. Johnson, R. D. Pierce, and W. J.
2 ~ Walsh, ANL Report on Contract W-31-109-eng-38, 1966.)
and in turbulent flow
Figure 15.8 illustrates the results of experiments where jM is plotted as a
5
function of Re for dissolution of .uranium spheres in flowing cadmium , cinnamic
(15.95)
acid spheres in flowing water, 2-naphthol spheres in flowing water 6 , and for
data are not quite as accurate in predicting mass transfer, although usually satis-
comparison purposes, jH 7 , and f /2 8 . Note that the data from the liquid metal factory to within an order of magnitude. .
experiments show lower values of jM at low Re values than those observed in the Also, many of the tests of applicability of heat transfer correlat10ns to mass
organic system experiments. This may be partially corrected for by the presence transfer involving the gas phase have not been made at conditions likely to be of
of the Schmidt number in the expression given in Eq. (15.96), which is not plotted interest to metallurgists, particularly at elevated temperatures and high rates of
in Fig. 15.8. No equation accounts for the peak in jM near Reynolds numbers of transfer, so they should be used with caution.
5
10 , so usually the graph is preferred rather than an equation in this range.
For natural convection, the mass transfer analog of
Example 15.4 Graphite particles are often added to molten cast iron in order to
(8.18) increase the carbon content when scrap steel is used as starting material. The time
required to dissolve the particles is of interest. Determine the time to dissolve
has been applied to both the rate of dissolution of carbon in molten iron 9 and the
10,
rate of dissolution of steel in molten iron-carbon alloys and found to be quite
particles of graphite as a function of the bath's carbon content.
The particles have a shape factor 2 of 1.5, and a characteristic dimension DP
_
reasonable as an approximation, with the best fit of the data yielding the equation of 0.14 cm. The particles float, and are swept to the side of the surface of the melt
ShL = 0.11 (GrM,LSc) 113 (15.97) by the magnetically induced stirring action resulting in a relative metal velocity of
approximately 25 cm/sec. Due to the displacement of metal by graphite, the
Many metallurgical processes depend on gas-liquid contact and mass transfer
surface area of an individual particle in contact with the metal is calculated as i of
between the phases. Although this is a very complex area, 11 several relationships the total particle surface area. In addition, a portion of the particle exposed to
have been found that describe the process of mass transfer on the liquid side of a the atmosphere is burned to CO due to the air circulating over the bath surface.
gas bubble-liquid interface. One of the most useful is that of Hughmark. 12 His Thus the recovery of carbon in the melt is less than 100 %, and experience shows
expression is
that, in fact, the recovery is only 50 %. Therefore, the mass of an individual particle
(of density p,) that dissolves in the melt is
kMd = [(V1 d)0.4B(~)0.339(g1/3d)o.on]b
D 2.0 +a v D v213 ' (15.98)
m=
l(n -
2 6DP 3 Ps ) '
where d is the bubble diameter, v; is the terminal velocity of the rising bubble,
and a and b are constants that depend on whether the bubbles act singly or in and its surface area exposed to the melt is
swarms. For single bubbles, a = 0.061 and b = 1.61. This relationship has been 3 - 2
A= 8 nDP 2.
tested in molten copper-carbon monoxide systems, and was found satisfactory
for describing the rate of transfer of oxygen in the copper to the interface. 13
Note that the Chilton-Colburn analogy is applicable only at relatively low ~ 0.1
mass-transfer rates, and that the best analogous results are obtained when similar
materials are utilized in both the heat- and analogous mass-transfer situations. ""'-~ 0.05
Turbulent flow mass-transfer correlations based on studies using common liquids
appear to be directly translatable into liquid metal systems, but heat transfer "0)
"<:)
E
0)
0
..."
6 ~
L. R. Steele and C. J. Geankoplis, AIChEJ 5, 178 (1959). ~ 0.01
7
T. K. Sherwood, Ind. Engr. Chem. 42, 2077 (1950). g
8 ~
0 005
9
M. Kosaka and S. Minowa, ibid. i - 0 3 4 5 678910 20 40 60 80100
10
M. Kosaka and S. Minowa, Tetsu-to-Hagane 53, 983 (1967). Peripheral velocity V, cm/sec
11
For reviews, see P. H. Calderbank, Trans. Inst. Chem. Engrs. 212, 209 (1967); F. H. H. Fig.15.9 Mass-transfer coeffic~s for carbon dissolution in Fe-C melts. (From R. G_. Olsson,
Valentin, Absorption in Gas-Liquid Dispersions, Spon Ltd, London, 1967.
12 V. Koump, and T. F. Perzak, Trans. AIME 236, 426 (1965); 0. Angeles, G. H. Geiger, and
G. A Hughmark, /. & E. C. Process Design and Development 6, 218 (1967).
13 C.R. Loper, Trans. AFS 76, 629 (1968); and M. Kosaka and S. Minowa, Trans. Japan Iron
C.R. Nanda and G. H. Geiger, Metallurgical Transactions 2, 1101 (1971).
Steel Inst. 8, 393 (1968).)
I
Solution. The mass flow, in terms of g C/sec, is the experimental data can be extended to the new situation. Several theories of
dm the process of mass-transfer have been developed, and they attempt to present
dt = - kMAPL(Co - C 00 ), models of what actually happens at the interface between two fluids or between a
fluid and solid from a fundamental viewpoint, in order to aid in intelligent extra-
where PL is liquid density, g/cm 3 , C 0 is weight fraction of carbon at the particle- polation of data.
melt interface, and C 00 = weight fraction of carbon in the bulk melt.
We evaluate the mass-transfer coefficient from Fig. 15.9 which has been lt1 ('
developed by several investigators for the dissolution of rotating carbon rods in
Fe-C melts. For a velocity of 25 cm/sec, kM for graphite is 0.02 cm/sec. Now,
f Phase 2
The oldest theory is the film theory of Lewis and Whitman 14 who suggested
that there is an unmixed layer or film in the fluid next to the actual interface,
continuously exposed to the completely mixed bulk fluid on one side and to the
other phase on the other. This layer, devoid of any fluid motion, is supposed to
or offer all the resistance to the transfer of component A from the interface into the
bulk solution, as depicted in Fig. 15.10. The transfer takes place purely by atomic
or molecular diffusion through the film. Figure 15.10 indicates the concentration
profile assumed in the model. Since the entire concentration change from CA 00 to
The results are plotted below. C1 is assumed to take place within the film in a steady-state manner, and since
the mass-transfer coefficient is defined by
15 .s
10 E
5 ~
"'
>
'"" 10
0 0.5 1.0 1.5 2.0 2.5 3.0
(Co-C00 ), wt
3.5
%
with the result that kM = D/i5.rr, where i5.rr is the effective film thickness. This
result often appears in the metallurgical literature in cases where the flux is mea-
sured and the overall concentration change (C1 - CA 00 ) is known, and then either
,-~
the diffusivity is known (or more often assumed) and i5.rr is calculated, or vice
15.9 MODELS OF THE MASS-TRANSFER COEFFICIENT ve11sa. As noted below, the significance of i5.rr is dubious, at best.
From a fluid mechanics standpoint, it was recognized at an early stage that
-s:ce most conditions in which mass transfer is important involve fluids under- interfaces between fluids are bound to be unstable with time, and that any given
going turbulent motion, we must usually rely on the preceding correlations, with element of fluid at the interface does not remain there for long. Thus, the film
experimental studies giving the necessary empirical coefficients. However, there
are many situations to which these test data do not directly apply, such as beyond
the experimental scope of variables, and it is desirable to know whether and how 14
W. K. Lewis and W. Whitman, Ind. Engr. Chem. 16, 1215 (1924).
- - - - - - - - - - - - - - ----------
theory is much too crude to be really meaningful. Higbie 15 proposed a model to elements (Higbie model) is long enough to allow the concentration gradient to
describe the contact between two fluids, in which he assumed that one fluid approach steady-state across the finite thickness of the element (film model).
exposes a "particle" of fluid to the other phase for an average time (}, which is This theory approaches each of the other theories as limiting cases. When D is
taken to be extremely short, such that the particle is subject only to unsteady-state large or the rate of surface renewal is small((} is large), then n approaches 1.0 and
diffusion or "penetration" by the transferred species during its contact time with so the film theory is applicable. When D is small or (} is small (rapid surface
the other phase. The particle is assumed to be stagnant internally during this time, renewal rate), n approaches 0.5, and the penetration theory results. In any case,
and well mixed before and after. Figure 15.11 gives a schematic picture of the there are still parameters that must be specified in order to use the theory for
situation. This theory results in a prediction that predictive purposes, and they are not readily obtainable.
kM = 2 fD
'1;e (15.99) PROBLEMS
~ At 1000F metal A is soluble in liquid B but B is not soluble in solid A as shown below
(QCA~+6C in{he pertinent part of the phase diagram.
cAro} Phase 2
Fluid element in
contact with phase 1
CAoo
Atom fraction B
Fig. 15.11 Schematic diagram offluid motion in penetration theory.
A 2-in. diameter cylinder of A is rotated at 1000 rpm in a large melt of 0.5 atom fraction B
The logical extension of this theory was performed by Danckwerts 16 who sug- at 1000F, and it is noted that after 15 min the bar diameter is 1.90 in. For the same temperature,
gested that the idea of a constant time of exposure (} ought to be replaced by an estimate the bar diameter after 15 min if another 2-in. diameter cylinder of A is rotated in a
large melt of 0.25 atom fraction B. We can assume that the molar volume of liquid A-B
average time of exposure calculated from an assumed distribution of residence
~is constant.
times of the "particles" at the surface. The result is again a relationship of the
form ~-2 )Use dimensional analysis to show
aJsh = f (Re, Sc) for forced convection;
(15.100) b) Sh= f(Gr, Sc) for natural convection.
15.3 Levitation melting is a means of supporting a metallic melt by an electromagnetic field.
The constants in his equation, like the constant (} in Higbie's equation, are not
No impurities are added in melting and operation under an inert atmosphere removes dis-
readily obtainable, with the exception of bubbles rising through a liquid in which solved gases. At 3000F and 1 atm hydrogen pressure, the solubility of hydrogen in iron is
case we may estimate (} to be the time required for a bubble to rise a distance 31 cm 3 per 100 g of iron. Estimate the rate at which hydrogen can be removed from a levitated
equal to its diameter. drop of iron that initially contains 10 ppm in the set-up shown below. Assume that no con-
When considering the two theories, it is apparent that the dependence of kM vection occurs within the iron drop.
on D is different. Experimentally, it has been found that kM is proportional to
D" where n varies from 0.5 to 1.0, depending on the fluids and the circumstances. O O Induction coils
In order to resolve this discrepancy, Dobbins 1 7 and Toor and Marchello 18 pro- 0 0
posed a combined.film penetration theory in which the residence time of the surface Levitated iron drop,
0 O ~-in. diam.
0 0
15
R. Higbie, Trans. AIChEJ 31, 365 (1935).
16
P. U. Danckwerts, Ind. Engr. Chem. 43, 1460 (1951).
17
I
W. E. Dobbins, Int. Conference on Water Pollution Research, London, 1962, Pergamon Vycor tube, t-in. diam.
Press, New York, 1964, page 61.
18
H. L. Toor and J.M. Marchello, AIChEJ 4, 97 (1958). Pure argon, 10 ft/sec
/
--~---~------------------------......----------------- .. - - - - - - - - - - - - - - - - - - - - - - - - -
15.4 Derive expressions for diffusion through a spherical shell that are analogous to Eq.
~5.1~(concentration profile) and Eq. 15.14 (molar flux).
A~)-Iydrogen gas is being absorbed from a gas in an experimental set-up shown in the
~below. The absorbing liquid is aluminum at 1400F which is falling in laminar flow
with an average velocity of 6 in. per min. 16
What is the hydrogen content of the aluminum leaving the tube ifit enters with no hydro-
gen? Assume that at T = 1400F, and 1 atm hydrogen pressure, the solubility of hydrogen
3
INTERPHASE MASS TRANSFER
is I cm per 100 g of aluminum, the density of Al = 2.5 g/cm 3 , and DH = I x 10-s cm2 sec. -1
-ltin.l-
~::::::;:::::::::::-...i-- Aluminum running O_:JO :
J, down the wall
'?ro\9\e<f;il<A_ \ ~- ~ In Chapter 15 we undertook much of the analytical presentation to familiarize
~ ourselves with mass-transfer coefficients as they arise from considerations of
-~ diffusion with convection in a single phase only. There are many situations,
however, in which two fluids or a solid and fluid are in contact, and interphase
~~, transfer by diffusion with convection takes place in one or both phases. In some
oases, convective mass transfer may be important in controlling the overall rate;
Surface concentration in others it is not important. If there is a reaction at the interface, it may very well
assumed to be equal to the be the controlling step.
saturation concentration
Experimentally, it is difficult to study interphase transfer, and at the same
time separate the individual phase resistances; so an overall transfer coefficient is
O:;D s'.,,\\o usually measured. Having determined this coefficient, we then attempt, via a
mathematical model, to deduce the individual phase coefficients, such as those
~ r1 UY'i ;f~l">U\ \ presented in Chapter 15. In some cases the results are clear cut, and we are then
able to adjust process conditions to optimize the process. However, in other
cases, we cannot differentiate between models even though they may be based on
important differences in basic assumptions, because their predictions can be
numerically similar within experimental error. Then, we deal with the overall
coefficients and utilize them, but only if the fluid conditions are similar in both
the prototype and model cases.
The objective of this chapter is to indicate the relationships between the
individual phase transfer coefficients and the overall coefficients, and to examine
several cases in detail, showing how different fluid conditions can influence the
overall coefficient through their effect on the individual coefficients.
Gas Liquid
-Interface
Fig. 16.2 The equilibrium curve for solute A partitioned between gas and liquid phases.
Distance temperature and partial pressure of A in the gas phase. On the curve, we identify
the four concentrations referred to in Eq. (16.1 ), specifically XA, Yl, X Aoo, and
Fig. 16.1 The two-resistance mass transfer concept.
YA 00 In addition, we show X AY and YAx which represent equilibrium concentra-
For example, assume that the only resistances to the overall process are tions corresponding to YAoo and XA 00 , respectively. YAx' in equilibrium with
diffusional resistances within the fluids themselves, and that there is no resistance X Aoo, is as good a measure of X A as X A itself, and moreover it is on the same basis
to the transfer of A across the interface. Consequently, the interface concentra- as y Aoo. We may then measure the entire two-phase mass-transfer effect in terms
tions, XA and Yl, are equilibrium values. In some instances, of course, the atoms of an overall mass-transfer coefficient, Kg:
cannot be accommodated into the new phase as rapidly as they might jump from
their original phase across the interface. If the accommodation process is slow (16.2)
relative to the diffusional processes, then XA and Yl would not represent equili- From the geometry of the figure. we can show that the relationship among the
brium concentrations. In most situations, XA and Yl are the equilibrium values, individual phase coefficients and the overall coefficient is
and we proceed with the "two-resistance" theory.
For steady-state transfer, the rate at which A reaches the interface from the NA/Kg= NAflM,g + m'NA/M,l'
gas must equal the rate at which it diffuses into the bulk liquid. Thus, if lM,g and or
liM,l are the local coefficients, then the flux of A is (16.3)
NA=liM,g(YAoo - YJ)= M,1(XA-XA 00 ). (16.1) In a similar manner, XAY is a measure of YA, and we may use it to define
2 another overall coefficient K 1 :
In this instance, since NA is given in moles/cm sec, and the concentrations are
expressed as mble fractions, the mass-transfer coefficients (liM,g and liM, 1) have the (16.4)
units of moles/cm 2 sec. The units of the mass-transfer coefficients, as presented
It readily follows that
in Chapter 15 and found within the Sherwood number, are cm/sec. Of course,
/i's are derivable from k; specifically, we have liA = kAC, recalling that C is the (16.5)
total molar concentration of the solution in question.
Equations (16.3) and (16.5) lead to the following relationships among the mass-
In experimental determinations of the rate of mass transfer, it is usually
transfer resistances:
possible to determine solute concentrations in the bulk fluids by sampling. But,
because the concentration boundary layers are usually extremely thin, it is physic- gas phase resistance 1/lM,g
(16.6)
ally impossible to approach the interface sufficiently closely to measure XA and total.resistance = 1/ Kg'
YJ. Under these circumstances, only the overall effect in terms of either XA 00 or and
YAoo can be determined.
Consider the equilibrium of solute A as partitioned between the gas phase liquid phase resistance = 1/ lM,l.
(16.7)
and a solvent, depicted in Fig. 16.2. The curve for the system is unique at fixed total resistance 1/K 1
/
I
----------------------------..-~---
If the individual phase coefficients, .M,g and iM,t have roughly the same relative values of the mass-transfer rates, we need to express the concentrations in
values, then we can demonstrate the importance of the shape of the equilibrium- mole fractions. In the liquid phase, we convert the solubility as follows
partition curve. If m' is small (equilibrium-partition curve is fairly flat and solute
A is very soluble in the liquid), Eq. (16.3) shows that the major resistance isreally 7.0 cm 3 H 2(STP) I liter H 2 1 g-mol H 2 63.5 g Cu I g-mol H
1/M,v and the rate of mass transfer is gas-phase controlled. XH = 100 g Cu 1000 cm 3 H 2 22.4 liters (STP)H 2 1 g-mol Cu 0.5.g-mol H2
Under such circumstances, even large increases in iM,t do not significantly
= 3.96 x 10- 4 .
change Kg, and efforts to increase the mass-transfer rate are best directed towards
decreasing the gas-phase resistance. Conversely, when m" is large (solute A is (Note: XH is the mole fraction of Hin the liquid.)
relatively insoluble in the liquid) and .M,g and .M,t are nearly equal, the major We can express the equilibrium constant for the reaction as
resistance is in the liquid. In these instances, efforts to bring about substantial
increases in the rate of mass transfer are best directed toward increasing the liquid XH
Keq= ~
coefficient lM.t
When overall coefficients are used in practice, they are frequently synthesized
where YH is the mole fraction of hydrogen in the gas; it follows that Keq =
through the correlations developed (such as those in Chapter 15) for the indivi-
3.96 x rn=- 4 at 1150C. Therefore, we can express the equilibrium partition of
dual coefficients of the phases in contact. It is important to recognize the limita-
hydrogen between the phases as
tions in this procedure. For example, the hydrodynamic circumstances must be
the same as those for which the correlations were developed. This is especially XH = 3.96 x 10- 4~.
true in the case of two fluids, where motion in one may influence motion in the
other. In other situations. absorption of surface-active substances at the interface
By referring to Fig. 16.2, we can estimate m' and m" with YHoo = 0 and XHoo =
3.96 x 10- 4 . In addition, Eq. (16.1) indicates that
may slow down reactions (chemical or physical) to such an extent that the estimated
overall coefficient in no way reflects the magnitudes of the individual phase YHoo - Yi'!~ Xi'! - XHoo
coefficients.
On the other hand, most experimental studies of reaction rates measure
since we are assuming lM,g = lM,t Therefore
' overall coefficients, from which attempts are often made to deduce individual Yi'!~ XHoo - x~ ~ 10- 4 ,
phase resistances. Again, if experimentally measured overall coefficients are to be
used in practice, the conditiom. with respect to .fluid motion and reactant supply ' - YJ2 - YHx ~ oo-4) - 1 ~ 104
must also be similar, unless it is clear that the controlling resistance of the overall m Xi'!-Xttoo - 0 - (4 x 10- 4 ) - '
-I
~------ --~------------
in processing. There are situations, howev_er, in which exceptions to this generaliza- m iron. We shall assume that only a single phase of iron exists. Therefore, the
tion should be made, as the two following examples demonstrate. initial and boundary conditions are -
C(x, 0) = C; (uniform), (16.13a)
16.2.1 Carburization of iron with surface reaction and diffusion as controlling
factors
ac
-::;--- (0, t) = 0, (16.13b)
ux -
If an iron plate is exposed to a CHcH 2 atmosphere, carburization occurs at the
surface according to and
(16.8)
ac r
(16.13c)
-;--(L, t) +-(Cs - Ce)= 0.
ux D
. ~ and consequently carbon diffuses into the plate. The objective here is to describe
the kinetics of carbon diffusion. Condition (16.13c) merely states that the amount of carbon furnished by the
It has been proposed 1 that the surface reaction proceeds at a rate given by surface reaction must equal the amount diffusing into the interior. Note that
Eq. (16.13c) is exactly the same condition existing at the surface of a solid losing
1 dnc Pett. 2-v heat to a surrounding environment at T1 , at a rate dependent on the heat 7transfer
-A-d = ri - v - -r2PH2 C,,
t PH2 (16.9) coefficient. In fact, Eq. (16.12) and its boundary conditions, Eqs. (16.13a), (16.13b),
where dncf dt is the amount of carbon in g-atoms taken up by the surface area A and (16.13c), are exactly the same as Eq. (9.66) and its boundary conditions, Eqs.
(9.67}-(9.6<;}). Hence, Eq. (9.85) and Fig. 9.11 are proper solutions to the problem
per unit time, r 1 and r2 are reaction rate constants, Pett. and PH 2 are the partial
at hand if we merely replace
pressures in the gas phase, presumed to be constant, Cs is the surface concentration
of carbon in g-atoms/cm 3 , and v is a number between zero and two, depending T- T1 C- Ce
on the details of the reaction mechanism. by
T; - Tf C; - Ce '
At equilibrium, dnc/dt = O; therefore, from Eq. (16.9) we can obtain the
equilibrium concentration of carbon at/L 2 by Dt/L 2 ,
and
C
e
= !J..Pctt
2
. (16.10)
r1 Ptt2 hL/k by rL/D.
Substitution ofEq. (16.10) into Eq. (16.9) gives
1 dnc 16.2.2 Transport in the gas phase and diffusion as controlling factors
A dt = r(Ce - C,), (16.11)
In this section,. we examine a situation in which the gas phase composition does
where not remain constant, because the reactfon rate at the surface of the solid is so
rapid that the gas cannot be replenished instantaneously by the incoming fresh
gas. As an example, suppose we wish to carburize the surface of low-carbon sheet
We can incorporate Eq. (16.11) into one of the boundary conditions for the steel with a CHcH 2 atmosphere. The steel is in the form of an open coil so that
diffusion of carbon in the plate. Consider the region 0 < x < L in which x = 0 the gas can flow between parallel layers as depicted in Fig. 16.3. We assume that
is the center of the plate, and x = L is the semithickness of the plate. -The appro- the thickness of the sheet in the y-direction is sufficiently large so that diffusion
priate differential equation is into a semi-infinite solid holds. However, there is a depletion of methane with
increasing x because it is consumed by the reaction, Eq. (16.8). Hence, the con-
ac
- = D - -2
a2 c (16.12) centration of carbon in the steel varies with x as well as with t (time).
ot ox To simplify the problem considerably, we may disregard the variation of
where C is the carbon concentration and D is the diffusion coefficient of carbon methane concentration normal to the flow direction; in the solid, the diffusion
of carbon in the x-direction is ignored since the diffusive component in the y-
1
C. Wagner, Notes from Course 3.63, MIT, Cambridge, Massachusetts, 1955. direction is so much greater.
I
-----------""""r-~~'=""- ...--.."".,,..----- --------------------------------------. - - - I u : z ., - , - - - - - . _ - - - -
......
Ce= C,(0)
\ '-C,(x)
- 2(bdx{D(~~L= 0 J
where b is the sheet width normal to the plane of Fig. 16.3, and C is given in
g-atom/cm 3 . Equating this value to the loss of methane per unit time, and sub-
stituting Eq. (16.16), we get
Fig. 16.3 Carburization of steel by gas flowing between parallel sheets. (~~L=O = 2b;:eq (~~L=O 6 8
(l .1 )
for x > 0. To summarize, we wish to solve Eq. (16.14), subject to Eqs. (16.15),
Therefore, Fick's second law for diffusion in the steel applies, and is written (16.17), and (16.18). A method of accomplishing this has been presented by
2
Wagner, and is qS follows.
oc
- = D -2,
02c Assuming that C depends only on x/Jt
and.y/Jt, we define the dimension-
(16.14)
ot oy less variables:
where C is the carbon concentration, and D is its diffusion coefficient in the steel. ~ = 2bDK0 q _x_. y
1J = 2fo.
Initially, the steel contains C; of carbon; hence nP 2fo'
C(x, y, 0) = C;. (16.15) Substitution of~ and 1J into Eq. (16.14) gives
Along the plane y = 0, we presume that equilibrium exists between the surface 2
o C 2(i: oC oC) = 0. (16.19)
carbon and the atmosphere according to 011 2 + " o~ + 11 011
K eq -_CSP~2
__, Substitution of~ and 1J into Eqs. (16.17), (16.15), and (16.18), respectively, yields
PcH.
C(~ = 0, 17 = 0) = Ce, (16.20a)
where K 0 q is the thermodynamic equilibrium constant for the reaction, Eq. (16.8).
C(~ = oo, 1J = oo) = C;, (16.20b)
Thus, we can represent the relationship between the local methane mole fraction,
XcH 4 , and the local surface concentration as
(16.20c)
Keq
C(x, 0, t ) = X CH.(x, t ) - = C., (16.16)
p We can transform Eq. (16.19) into an ordinary differential equation if we
consider Ca function of the sum a = ~ + 17.
where X CH is the mole fraction of methane, P is the total pressure, and X CH I.
At the entrance, we have the original mole fraction, xgH; thus d2C 2a da = 0. (16.21)
da 2 + dC
C(O, 0, t) = xgtt 4
Keq = Ce,
p (16.17) The solution to Eq. (16.21) subject to Eqs. (16.20a, b, and c) is
where Ce equals the equilibrium concentration for the original gas composition. 2bDK0 q ]
As the gas flows between the parallel plates from x to x + dx, the loss of
C-C. nP x+y
---'-' = 1 - erf [ (16.22)
methane per unit time equals Ce - C; - 2fo
- n( ~~H)dx,
0 Example 16.2 Calculate the distance x over which the surface concentratfo;Cs
does not vary excessively after carburizing. Specifically, it is required that the ratio
where oXCH)ox is negative, and n is the flow rate of the gas in mol/sec. The 2
C. Wagner, ibid., and Zeitsch. f. physik. Chem. 192, 157 (1943).
I
~----~"->U-In~pn~1:11:n~-1."J1""G"1J.,.11-.;;---- ---------------------------
(C; - C;)/(Ce - C;) ;;::: 0.75. Surface carburization is carried out with dilute 16.2.3 Mixed control in oxide reduction reactions
methane/argon-hydrogen at 1000C and 1 atm, and the appropriate equilibrium The rates of reduction reactions are of practical interest to metallurgists, and many
constant is 0.7 atm g-atom/cm 3 . The gas flows at a linear velocity of 10 cm/sec studies on this subject have been reported in the literature. In most cases, some
through a separation of 2 cm between steel sheets for 5 hr. At 1000C, the diffusion attempt has been made either to ascertain the rate-controlling step or mass-
coefficient of carbon in y-iron is 3 x 10- 7 cm 2 /sec. transfercoefficient for a specific step in the overall process. However, what is usually
measured is the overall coefficient, and in many instances, unless conditions are
Solution. Seeking the surface concentration C,, we obtain from Eq. (16.22)
clearly such that one step does not control and only one other step is involved, it is
very difficult to be sure of the rate-limiting step.
[
2b~Keq X] As a classic example, consider the reduction of iron oxides by gases, which has
Cs - C; l f nP been the object of scores of papers. The measurements made are usually those of
Ce - C; = - er 2-JDt .
weight loss of a specimen as a function of time. If we assume that the overall rate is
We require the left-hand side of the equation to be 0.75. Hence controlled by tpe rate of the chemical reaction
FeO + CO(g) =Fe+ COi{g)
0.25 = erf [
2bDKeq ]
2
fo x .
at the oxide-metal interface, then, for a spherical particle of initial radius r0 and
density p 0 ,
(16.23)
We find the argument of the error function to be 0.225, so that
should relate the fractional reduction f to time t. From the plots of test data, we
obtain the values of the rate constant kc, and plot them as In kc versus y- i in order
to obtain activation energies for the process. On the basis of agreement of test data
with Eq~ (16.23), such as in Fig. 16.4, one might be tempted to conclude that the rate
We can substitute all values directly, except for n/2b which is evaluated simply as of reduction of iron oxide is controlled by the rate of the chemical reaction step.
However, it happens that in another study the data fit another model that presumes
2
n = 10 cm b 2 cm 273K
---+----+---~+----'---~--
1 g-mol that the rate of reduction is entirely controlled by equimolar countercurrent
sec 1273K 22,400 cm 3 (STP) diffusion of the reactant gas and product gas through the porous metal product
layer that forms. These data are shown in Fig. 16.5. Another study of data based
Therefore on many investigations again concluds that the rate is controlled by the chemical
n (10)(273)
2b = (22,400)(1273) g-mol/sec-cm.
-~ 0.6
7 2
~
Then, by substituting t = 18,000 sec, D = 3 x 10- cm /sec, P = 1 atm, and ::, 0.5 pure CO
..!_.
Keq = 0.7 atm g-atom/cm 3 , we calculate x to be 15 cm. 0 0.4
Q.
Of practical significance, this section shows that excessive nonuniformity of ~
0.3
the surface concen~ration can only be prevented by a sufficient concentration of Size I mm
2mm
sphere O
sphere
the reactant species in the exit gas phase. For example, we saw .in the above 0.2 3mm sphere .A.
4mm sphere.
problem that there was a limit to the distance that could be properly carburized 0.1
for the specifications put forth. If the specifications had been more severe, for
0
example, [Cs - C;/Ce - CJ ;;::: 0.90, then a higher velocity of carburizing gas 0 10 20 30 40 50 60 70 80
would have had to be supplied to achieve the same value of x, thus raising the t,min
concentration of the reactant species in the exit gas. If the active species is valuable, Fig.16.4 Experimental data for hematite reduction by CO gas plotted according to Eq. (16.23),
recovery of this component or recirculation of the exit gas must be used. This is an which is based on the assumption of chemical reaction control at the oxide-metal interface. __ .
important economic consideration in chromizing steel with CrC1 2 (g). (From W. M. McKewa~, Trans. AIME 212, 791 (1958).)
I
oxide-metal interface itself. If the mass transfer and diffusion steps are considered
together, the mass flow to the interface from the bulk gas stream is
(16.24)
t Product
layer
t reaction interface. When we compare the experimental rate of weight loss, Re,
to the rate of weight loss predicted by Eqs. (16.24) and (16.2.5), we see from Fig. 16.7
that the ratio RelRn increases rapidly toward 1.0 as the thickness of the porous iron
t t q =C*
layer increases. Thus, it is apparent that there is an interaction between the mass
transfer steps on the one hand, and the chemical reaction step on the other, with
the latter contributing to the control of the overall rate only at early stages of the
t t
C= C00
reduction process at high temperatures, or at lower temperatures generally.
t t ~
Cl
1.0
""o
1000 C
t Ct::~
~
0.8
Fig. 16.6 Schematic model for oxide reduction.
~
0.6
reaction at the metal-oxide interface, 3 but in this case an entirely different activa-
tion energy than in the previous work is obtained.
0.4
More recently, Olsson and McKewan 4 have shown that in fact there exists %H2
mixed control. Their model, as defined in Fig. 16.6, assumes that there are three 0 100
steps in the overall process: gas-phase mass transfer to and from the bulk gas 0.2 /J. 90
D 80
stream and the solid, diffusion of the reactant and product gases across the solid,
but porous, reaction product layer, and, finally, the chemical reaction at the
0 0.05 0.10 0.15 0.20
Iron layer thickness, cm
3
N. J. Themelis and W. H. Gauvin, Trans. AIME 227, 290 (1963). Fig. 16.7 The ratio of the experimental to the calculated diffusion limited rate of weight loss.
4
R. G. Olsson and W. M. McKewan, Trans. AIME 236, 1518 (1966). (From R. G. Olsson and W. M. McKewan, Trans. AIME236, 1518(1966).)
l
-------------------------------~ -- --- - -- - - --- -- --- - - - - - - - - - - - - - - - - - - - . , . , . - - - - - - - -
Wen 5 has discussed in detail the mathematical solutions to the kinetic equa- Table 16.l Clausing factors for effusion cells*
tions for many different models of gas-solid reactions; he has shown how similar
some of the results are in terms of experimental weight loss versus time data l/r w.,
for models differing widely in terms of assumptions made. From his paper and the
0 1.00
example above, it is clear that misleading conclusions concerning rate-controlling
0.1 0.9524
mechanisms can easily be made, unless much care is taken to ensure that experi- 0.2 0.9092
mental conditions eliminate control by steps being ignored in the analysis. How- 0.3 0.8699
ever, it should be kept in mind that overall coefficients can be useful information 0.4 0.8341
from the design engineering standpoint.
* From P. Clausing, Annalen der Physik 12, 976
(1932).
16.3 MASS TRANSFER WITH VAPORIZATION These expressions give the net rate at which atoms or molecules leave the cell
through lli'rinfinitely thin orifice opening. Because of the difficulty of constructing
As pointed out in Chapter 5, there are several situations in metallurgy that involve
heating materials in a vacuum. In vacuum melting or heat treating of ferrous an infinitely thin opening, that is, with no walls, there is a correction to the theoretical
alloys, valuable elements, such as manganese or chromium, may be lost, or con- flux accounting for the fact that there is a small probability of molecules rebounding
versely undesirable impurities, such as zinc or lead, may be removed. Experi- and not passing through the opening. The Clausing factor, We, gives the probability
mentally, in Knudsen effusion cells, use is made of the measured rate of vaporization of passing and is a function of the ratio of the orifice thickness to radius, l/r, as given
of an element to determine thermodynamic properties of the material. in Table 16.1. When this factor is included, then
In all of these cases, there are several steps involved in the overall mass transfer AeW.,(P1 - P2)
process: 1. transport of the volatile species to the surface of the condensed phase; J = . (16.26)
J2nMRT
2. formation of volatile compounds at the surface; 3. evaporation from the surface
into the gas or vacuum; and 4. transport away from the surface into the gas or At the same time as vapor leaves through the orifice, the solid within the cell
vacuum. Depending on conditions, any one of these steps may be rate limiting, or loses alloy from its surface initially having an alloy concentration of CJ. If the area
several may comprise a mixed-control process. of the orifice, Ae, is small relative to the exposed surface area of the alloy, A., then
we may take the partial pressure within the cell, P 1 , as uniform. Furthermore, P 1
16.3.l Knudsen effusion cells
is the pressure of the volatile component in equilibrium with its concentration on
The Knudsen cell is used to study thermodynamic properties of alloys by taking the surface of the solid, CA. Therefore, Fig. 16.8 gives.the situation, and the diffusive
advantage of the high volatility of one component relative to the others, and , flux within the solid to the surface equals the effusive flux
determining the vapor pressure of this species as a function of alloy content. The
vapor pressure is determined by measuring the force exerted by a molecular stream J = + DA.(acA) AeW.,(P1 - P2)
J2nMRT '
(16.27)
emanating into a vacuum from a very small hole in the end of a hollow cell con- OX x=O
taining the alloy. Under these conditions, free molecular diffusion occurs; then,
where CA is in mol/cm 3 .
by referring to Section 5.4.1, we have
znet = vr;z;;T
2-;;:; (n1 -
2
nz), molecules/cm -sec, p 2 (vacuum)
Source~
alloy
or
Ae(P1 - P2 ) coA
A,
A, __..
J
5
J =
/
y2nMRT
, moles/sec,
vapor pressure of the pure component at the temperature in question. Thus, with
P 2 = 0 in a vacuum, 6
Ae W.,yAPJCAs
D (acA) (16.28)
As OX x=O pj2nMRT 4
3
where p = mol alloy/cm , the molar density of the alloy. Fick's second law
describes diffusion within the solid; the boundary condition at the surface of the
solid is Eq. (16.28).
- 2
-CA = erf
CJ
rn: + exp (Yx + Y Dt). erfc ( 2vxrn:Dt + Y v'rn:)
x
2v Dt
2
Dt . (16.31)
the particular physical arrangement and stirring conditions involved. If there
is no fluid motion, the problem reduces to the same conditions as in Section 16.3.1.
In the case of induction melting, there is no general correlation available, so we
The surface concentration of the solid Cl, obtained by letting x = 0 in Eq. shall make use of an hypothesis by Machlin, 8 which states that flow across the
(16.31), is surface of an inductively stirred melt may be considered to be a free streamline slug
flow without any shear gradients taking place within the surface layer. The
CjjCJ = exp (Y 2 Dt) erfc (Y fti). (16.32) volatile solute is supplied to the surface solely by diffusion from within this surface
7
Using Eq. (16.32), Schroeder has calculated the maximum experimental times layer. This hyp6thesis is exact!)'- the same as Higbie's "penetration theory"
discussed in Section 15.9. The only new assumption is that the "lifetime"() of a
6
H. S. Carslaw and J.C. Jaeger, Conduction of Heat in Solids, second edition, Oxford, 1959, surface element is equal to the average distance from the center of the melt to the
page 71.
7 8
D. L. Schroeder, Sc.D. Thesis, MIT, 1966; also Trans. AIME 236, 1091 (1967). E. S. Machlin, Trans. AIME 218, 314 (1960).
l
-----------------------------------------------------------------------
edge of the crucible, divided by the average velocity of the melt at the surface. In where K is the overall rate constant,* defined as:
the induction stirring case, {) is of the order of 1 sec or less.
For this step in the process then, the average mass-transfer coefficient is 1/K = ljkM,l + ljkM,e
(D -2
Solving for Cl, we get kM,l = 2'.}-;e = 1.08 x 10 cm/sec.
l
or
K = 8.8 x 10- 3 cm/sec.
0.8 1
It is apparent that when there is a good vacuum, liquid phase mass transfer offers IO
81 % of the total resistance; thus we may say that it essentially controls the rate. 0.6
This changes, however, if the stirring is increased to decrease 8. For example,
if e decreases to 0.1 sec, then 0.5
Now the liquid phase mass transfer and the evaporation process are almost
equivalent in their contribution to K, and there is mixed control. Furthermore, K
has increased to 1.97 x 10- 2 cm/sec.
- 0.2
'
Pressure, a tm
10-6 10-s 10- 10-3 10-2 10-1 I t,min
~ 10- 1
8u Fig. 16.11 Decrease in concentration of alloying element as a function of mass-transfer
~ Io- 2 coefficient and A,/V ratio.
10-3 Some other experimental values of K for the loss of various elements from
iron-base melts into good vacuums are given in Table 16.2.
10-
16.4 THE EFFECT OF TEMPERATURE AND THE CONCEPT OF
THERMAL STABILITY
10-s
1.0 IO I0 2 I0 3 10 I0 5 I0 6
.
For any reaction between a gas and a solid, there are several temperature regions
Pressure, m
where control of the reaction may be by different mechanisms. In the low tempera-
Fig. 16.10 Overall mass-transfer coefficients for Mn vaporization from 0.25 % C steel at ture region, the chemical reaction is often the slow step and controls the overall
1580C asa function of the pressure of Aroverthemelt. (From R. G. Ward,JJSJ201, 11 (1963).) rate. The rate of change of the chemical reaction rate with temperature is exponen-
tial, since in general
Since neither kM,t nor kM,e depend on the external pressure, K is independent kc= Zexp(-t1H/RT), (16.38)
of vacuum pressure as long as kM,g remains very large. When kM,g becomes about
Table 16.2 Overall evaporation constants for iron-base melts at 1600C*
an order of magnitude smaller than kM,t or kM,e, it is rate controlling. Since kM,g
is proportional to p- 1 , then K decreases with P, as long as b.rr in the gas phase Element K,cm/sec
remains constant. When the gas phase is dense, then kM,g is controlled by natural
or forced convection. Figure 16.10 illustrates the effect of system pressure on K for Mn 8.4 x 10- 3
vaporization of manganese from steel. Figure 16.11 gives the results of the applica- Cu 4.8 x 10- 3
tion of Eq. (16.37) to calculate the rate of loss of manganese at 1600C for various Sn 2.3 x 10- 3
(A 5 /V) values, using K = 0.01 and 0.001. One apparent result is that the use of an Cr 2.1 x 10- 4
argon atmosphere to increase pressure decreases K to the point where the loss of s 7.0 x 10- 4
manganese is negligible.
* From R. Ohno and T. Ishida, J. Iron and Steel Inst., London 206, 904 (1968).
I
-------1ne-errecrorremperamre-an1rme-conceprormermarsraumry--569
Thus, as temperature increases, the chemical reaction rate usually increases rapidly Remember that kc increases exponentially with temperature, whereas D varies
to the extent that it no longer is the slow step, and diffusion of either the reactants approximately as T 2 for gaseous diffusion. Therefore, at low temperatures,
or products to or away from the reaction surface becomes the slow step. This is kc D/berr, and
depicted in Fig. 16.12, where, as the temperature increases, the proportion of control
by diffusion increases. j ~ kcCoo.
On the other hand, at high temperatures, kc D/bem and
D
j ~~Coo.
Ueff
These are the limiting cases; and in between, the control is mixed. The absolute
Chemical value of the temperature where this transition occurs depends, of course, on the
reaction
control parameters involved in each particular case. As an example, take Fig. 16.7 for the
Temperature
case of hematite reduction. The upper limit of mixed control in this case appears
to be l000C, and the lower limit is at some temperature less than 800C.
Fig. 16.12 Schematic reaction rate as a function of temperature for a gas-solid reaction.
If we compare the reaction rate dependence on temperature with the rate of
heat loss from a given system, ~!so as a function of temperature, we can obtain a
Consider a simple model in which the rate of the reaction is proportional to the semiquantitative picture of what is involved in the thermal stability of a process.
concentration Cs of the gaseous reactant at the reaction surface: The shape of the reaction rate curve is important because the rate of heat release
(in an exothermic reaction) is directly proportional to the reaction rate. Consider
(16.39) the situation in a packed bed reactor. If the gas entering the bed is at ~,and the
The rate of diffusion of the reactant from the turbulent bulk gas stream of con- solids are at a uniform temperature T,, then tne rate of heat loss per unit volume
centration C00 to the reaction site is given simply by from the solids (which are presumed to be overheated by the evolution of heat of
reaction on or within them) is
(16.40) Qloss = hS(T, - ~). (16.42)
Therefore, the rate of heat loss from the solids is a linear function of T,, for a given
where <>err is the effective thickness of the fluid through which reactants diffuse to
the reaction site. ~,as shown in Fig. 16.13 for several values of hS. The heat evolution curve, on the
other hand, follows the same shape as the reaction rate curve of Fig. 16.12.
Stable steady-state conditions occur when the two curves cross. Note, however,
that in the cases of multiple intersections, such as where the rates of heat addition
and loss are equal at point C, any slight increase in temperature results in more heat
being generated than removed, causing the temperature to rise to point D. Any
further temperature increase results in more heat being lost than generated, thus
returning the system to D. A similar line of reasoning shows that a slight decrease
B from point C reduces the reaction temperature down to point B. Therefore, in the
Temperature
case with a heat loss rate of Q*, there are two stable states, one corresponding to the
slow, chemical-reaction-con trolled kinetics, and one to the fast, diffusion-controlled
Fig. 16.13 Interaction of heat loss and heat evolution curves for various heat loss conditions.
kinetics.
Con'i~enng t~ra~~rt~e~thern ~ '~g~ '' po,,ible ,iruation' ro, a
hS value, depending only on the gas temperature, as indicated in Fig. 16.14. In the
~~ ~-- lliribefue~nd.,ion ch~t~ "~~fmelting~h~ ru~~:;q~
'
b) that
is carbon-saturated pig iron at 2600F.
fue the
case where~ = Tg~ the stable steady-state conditions are those described above. c) Repeat (b) when the bath temperature is 2900F and the carbon content is 0.20% C.
However, ifthe entering gas temperature is decreased, the heat loss curve shifts to
the left until a situation is encountered such as that for ~ = Tg. If the process
operates at point D, it decreases in rate and temperature until point X has been
reached. At that point, any further slight decrease in process temperature greatly
decreases the rate of reaction, because the system rapidly approaches point Y.
On the other hand, an increase in entering gas temperature from ~*to ~**
results in slight increases in rate and process temperature if the process is initially
.' ~,at either point B or point D, and in the possibility of rapid and uncontrollable
temperature rise from point Z, once it has been reached. ~** is known as the
minimum gas temperature for ignition of cold solids, and the minimum ignition
temperature of the solids is T2 . By the same token, Tx is the minimum combustion
temperature or critical extinction point. These temperatures depend on the shapes
of the heat loss and heat evolution curves, which in turn depend on the gas flow
rates, surface areas, heat-transfer coefficients, diffusivities, gas compositions, etc.
Thus, there are certain regions of operation of gas-solid systems in which stable
operation is impossible.
Fig. 16.14 The effect of variation in r:i on thermal stability for a constant value of hS.
PROBLEMS
16.1 Derive Eq. (16.24) leading to the diffusive and mass transfer flux to the reaction interface.
~Estimate the rate of loss of Zn from 70-30 brass by evaporation alone as a function of
~erature.
16.3 Some ferritic stainless steels are vacuum heat-treated in order to maintain surface finish.
If AISI Type 410 parts are heat-treated in a furnace with an exhaust duct area of 200 cm 2
and a vacuum of 10 of Hg at 1600F for 2 hr and their surface area is 2000 cm 2 , what will the
surface concentration of chromium of the parts be? Assume that the initial concentration is
12 %Cr and De, = 10- 12 cm 2 sec.
16.4 The basic oxygen process of making steel involves melting some steel scrap in the
furnace. Occasionally, large pieces of scrap remain unmelted at the end of a heat.
a) Which is more important in determining the rate of scrap melting, the weight of the piece
of scrap or its dimensions?
APPENDIX I
*Note: Throughout this book, the term calorie, unless otherwise qualified, means thermo-
chemical calorie. It is arbitrarily defined by the relationship: 1 thermochemical calorie =
4.184 absolute joules, or 4.193 international joules.
The absolute joule is defined by the equation: 1 absolute joule = 10 7 ergs. One erg is
the work done when a force of one dyne acts through a distance of one centimeter. One
absolute or mechanical watt is the rate of working of one absolute joule per second.
The British thermal unit and the International calorie are in common industrial use.
They are defined by the relationship: 1 international calorie/gm = 1 Btu/lb when 1 inter-
national calorie = 4.186 international joules.
.t 573
J
,:;>
2 *
:i.>
::I 0..
(TQ"' Table Il.1 Densities and thermal properties of various substances
n'S
0"'
' 0..
("):::;> Spccific heat in Btu/IbF (or gm cal/gmC) J Meltin Boilin~ Latent ;eat
.,n
Alcohol, erhvl (50%) and warer Liquid ......
Alcohol, erh~l (10%) and wacer s
S7.3 0.923
::ii
~
~
Liquid 61.4 0.99
A:cohol, methyl (CH.JbH) Liquid 0, F 49.6 0.601 S9-6S
~
0.4580 214-133 -142.6 ISO.B 29.S 480.6
Alcohol, methyl (90%) and water Liquid s Sl.4 ~ 643 ......
<:; Alcohol, methyl (503) and waur Liquid s 57.3 0.846
::s Alcohol, methyl (10%) and water Liquid s 61.4
0.9861 ,_J
~
Alumina-see Aluminum Oxide, Alumina
(fused) refractory, high-alumina refracrory ~
::s
~
Alumina (fused) refractory Solid_ R- - -l~HBl . -0:20 60-1200 3390+ .... . ...
:: (see also high-alumina refractory)
"" Aluminum;_ - So!id M, E _ 166.7
i32S 61-S79 1220 3272 167.S 360.0 -.!
'""
'O
Aluminum foil
Aluminum oxide (alumina)
Soli<l
Solid
r
243.S
0 4
0.183
290
32-212 ..... 366B '"' ...... . ..
~
~
Ammonia (NH3) Gas 0.046 1.08 -B.4-+s 0.S2S 70-220 -103 -28.3 lSO SB9
<= Ammonium chloride (10'-);J anrl water Liquid s 64.4 0.778 ......
~ Ammomum sulfate [(NH:.i)2 SO.i] Solid 110 284 9s6b ..... ""'!
::I"
8.
Andalu~1te (Al2 SiO.~)
Aniline (Ci;H7N)
Antimony (Sb)
Solid
Liquid
Sol:d
0, F
M, E
199.8
63.9
422 1
0.228
O.OS2 392
O.Sl4 S9
3290
17.6
1166
363
2624
3B
6B.9
19B.O
703
'
~ Argon (A)
Arsenic, grar
Gas
Soid
E
E
0.105
35B 0.0822 32-212
0.1233 68-194 -306.4
c
-303
c
12 68
"'
::I a-A=alloy, E=clement, F=fucl or fuel component, I= insulation, M=mctal, O=org3.nic compound, R=refracrory, S=solucion.
b---decompo!'es.
~
C-<.ublimes ar 1038F, mel [Sat 1562F unde.r 36 atmospheres of pressure.
"'<=
::I
iii'
7
(fusion~
in Beu/lb
Normal I Othcra Density Solid stare L1q111d state I Gaseous state point, F
1n pomt (at std
Name - Description state classi
fl.cation lb/cu ft Specific
h~at
Temri
range F
Specific I Temp
heat range F
Specific I Temp
hr"at range F
F barometric
pressure)
I of
fusion
I of vapor
1zat1on
~1--A--1--.-- ~ 60-462 --;-;;-i---1---i---~---.-~---
Babbitt, lead base 1
tin base J Sohd A 46S 0 071 60-464 0 063 I 464 34.1 , .. .
Cadmium (Cd) So: id E, M S40 0.057 212 609.6 1412 23.B 409
Calcium (Ca) Solid E 96.6 0.170 32-3SB 1S60 262S
Calcium carbonate (CaC03) Solid 168-IB4 0.210 32-212 1s17b
Calcium chloride (CaCl2) Solid 134 0.292 60 ' 142).2 ,2910
I
s
Calcium chloride (30o/0 ) and water
Camphor (C10H60)
Liquid
Solid 0
7B.7
62.4 I o.44 68-3S3
0.676
0.61 I j5j 408.2 19.4
:t-A=a11oy, E=clemenr, F=fuel or fuel component, !=insulation, M=mecal, O=organic compound, R=refractory, S=solucion.
b-decomposes.
",
.
I
Normal Orhe.ra De~sicy
I
Salid
spcci"fi c heat m. Btu,/lboF
~t:!Ce
, ( or gm ca I/ gm oc)
l_iqmd state Gaseous state
I(fus~on). I
I
Mel~m~ Boilin~
point, F
I Latent
B heat
m tu
/lb
Copper Sulfate (CuS04) (16.7%) and water Liquid S 73.7 0.848 53.6-59
Cork, natural Solid I 15 0.419 77
Cork, granulated Solid I S.4-7. 3 0.43 77
a-A=alloy, E=clc:mc:nt, F=fucl or fuc:l component, I= insulation, M=mc:tal, O=organic compound, R=refractory, S=solution.
C-sublimes.
.,,__,.,,,,.""""""_ _ _ _ _!1111111111_ _ _ _ _ _ _ _ _ _ _ _ __ _
-
~~~~~~~~--- ~~~---.~ ---~-
Mame - Description
I Normal
state:
Othc:ra
classi-
Dc:nsity
in
Specific heat in Btu/lbF (or gm cal/gmC)
Solid state Liquid uace Gasc:ous srate
Mc:ltin~
(fusion
point
Boiling
point, F
(at std
Latent hc:at
in Btu/lb
I
fication lb/cu ft Specific Tc: mp
heat range F stc:'~~c Temp
range F
Specific
bear
Temp
range: F
OF baromc:cric
pressure)
of
fusion
of Vafor-
--- izacion
Corkboard Solid ' I 6.9-20.7 0.204 -115
----
'' ...... " "
0.417 +1so "
"-A= alloy, E=elc:ment, F=fuc:I or fuel componc:nr, I= insulation, M=mc:ral, O=organic compound, R=refraccory, S=solurion.
Table 11.1 Densities and thermal properties of various substances (continued)
Galena (PbS)
Gallium (Ga)
So!id
Solid E
467
367
0.0466
O.OEO
32-212
54-235
' I 2050
86.18 3090
Gasoline (commercial) Liquid F 41-43 0.5135 158-194 128-146
Germanium (Ge) Solid E 335 0 0737 32-212 1756.4
German silver Solid A ( 0 0946 32-212 \ 0.123 1850 86.2
\ 0.109 60-1850 f
Glass Solid I
144-187 0.15-0.23 32-212 2190
I 0.132 -117 i
Glass block, expanded, foamglas Solid 10.6 \ 0.157 - 20 I
I! o 179 112 \I I I
Glass wool Solid I 1.4-4.8 I o 210 150
'\ 0.279 624 J
Glycerine (C3Hg03) (glycerol) Liquid 0 78.7 0 576159 212 I -68 554 76.5
Gneiss Solid 0.196 63-210
Gold (Au) Solid E, M 1205 0.0316 32-212 0.034 1945.4 5380 28.7 29
Granite Solid 162175 0.192 54-212 ...
Graphite Solid 138.3 r 0.201 32-212 \ 6300 8720
\ 0.38 702200 (
Gypsum Solid I 145 0.259 )0-212 2480
r- ------~.~
~.--~
,,-~~o . .....,,.<;>,..:";-.,..,.,.., 0, ''""""!~ 4-!"<" ' ~r""-4~_m"!4Ji3 JJ,tl!&&
J
'1
0.1069
65
a.so I 32-212
-272.2 -241
I ias:110
Krypton (Kr)
I .... )I ....
1145-15~
Lead slag wool Sol;d I J 0.178 150
\ 0.235 718
Light oil Liquid F so 0.50 ....
Limestone Solid 168-175 0.216 59-212
60-486 486 21.S
t~~~::1~il
Solid A 0.036
Liquid 58 0.441 -s 600
Lipowitz's metal (Pb26, Snl3, CdlO, B151) So Ed A 0.041 60-140 140 17.2
Lithium (Li) Solid E 33 1.0407 212 366.8 2552 286
Lodestone (magnetite) Solid ..... 322 0.156 32-212
a-A= alloy, E=dement, F=fuel or fuel component, I= insulation, M=rnetal, O=organic compound, R=refractory, S=solutioo.
~;
Name - Description
Normal
state
I Othe~a
class1
Dc:~sity Solid state Liquid state [ Gaseous state (fusion
point
point, 0 p
(at std
ficacioa
tn
lb/cu ft Specific
heat
Temp
range F
Specific
heat
Temp
range: 0 p
Specific
heat
Temp
range F
F bacommicl
pressure)
of
fusion
I of vapor
ization
---
Solid
--- - - - -0.27
- - --- - - - - - - - - - - - - - - - - - - - - ---- ----
...
Magnesite refractory 60-1200
R 171 mo+
Magne~ite refractory (unburned) Solid R 183 0.26 601200 .. Jsso+
Magnesite refractory (fused) Solid R 179 0.27 60-1200 ..... JS80+
Magaesiu'." (Mg) Solid M,E 108.6 0.2492 68-212 0.266 120J.8 12048 I 160
Magnesium oxide Solid 228 ( 0.2J4 86-100 \ ... 5070 6S80
0.29S 86-1800
Magnetite (Lodestone) Solid J22 O.IS6 J2212
Manganese (Mn) Solid E,M 464 0.1211 68-212 ..... . .. .. 2246.0 J4S2 I 11S
Marble Solid !62-17S 0.210 J2212 ... . ...
Meccuric chloride (HgCl2) Solid JJ9 .. .... . ... SJ9.6 S81
Mercurous chloride (Hg2CJ2)
Mercury (Hg)
Solid
Liquid E, M
446
847
o.os
O.OJ2 1-m. \
+68
68
J
0.0JJ Jl-212 ... . ...
S76
-J7.97
722
674.6 I S.07 I 117.0
Mechane (CH4) Gas 0, F 0.0424J ..... 0.992 -172 o.s929 64-406 -260
Mechyl chloride (CH3CI) Gas 0 .!JS O.J8S 60 ..... -144 -11
Mica Solid 0.10 68
Mica, expanded (vermiculite) Solid I ... ( 0.20S 149
0.2J6 290
Milk Liquid .. 64.2 ... I 0.847
Mineral wool board with binder Solid I 14.J ( 0.09S -120
0.2'\7 ISO
Molasses Liquid 87.l 0.60
Molybdenum (Mo) Solid E,M 6J7 0.0647 68-212 ..... 4S9S 126
Monel metal Solid A 550 0.129 60241S 0.139 241S 16680 117.4
Naphtha
Naphchalene (C 0H4C4H4)
Neac's-fooc oil
Liquid
Solid
Liquid
0
41.2
71.8 0.J25
0.4S7
68-140
6886
0.49J
0.427 262 .. l7S.8
J2
J06
424.2 I 64.1 I 184
13S.7
1~:::.2.
Neon (Ne) Gas E O.OSl4 0.44J any -42J'
Nichrome Solid A Sl7 0.111 ..
Nickel (Ni) Solid E,M SS6 J 0.109 64-212) 264S I 13J I 2670
\ O.IC08 1832
Nickel steel Solid A ..... 0.109 I >2-212
Nicric acid (HNU:J) Liquid 96.1
I
o.445 I I I I -4J.6 I 186.8 I I 207.2
NJCric acid (IO(/~) and water Liquid s I i 0.768
N'1tric and (2(,;) and water Liqu:d s I 0.930
Nicric acid(!<,,~) and water Liquid , s I I I 0.96J
1-A=alloy, E=clcmcnr, F=fuel or fuel componcnr, T=msularion, M=mcrll, O=organic compound, R=refraccory, S=solurion.
-------------"""'""'"""'~"""' . ." " " " " ' ' " -!. l'g~)~a.1wz.ms 4 rr
Name - Description
Normal
state
Other a
class1-
Denstty
Jn
Solid state
Specific Temp
I Liquid state
Specific Temp
Gaseous stace
Specific Temp
Melt in~
(fusion
p<?mt
OF
Boiling
pomt, F
(ac std
in Beu/lb
ficauon lb/cu fc
heac range F
---- ---- - - - - - - --- ----
heac Irange F heac range F
barometric
pressure)
of
fusion
------
of vapor-
izat1on
Nitric oxide (NO) Gas 0.079 O.S80 ( -249 \ 0.2Jl7 SSJ42 -240
-2S2 f
Ni crobenzene ( C5H502N) Liquid 0 74.7 O.J8 122 . .. . .... 41 411.8
Nitrobenzole Liquid O.JS S7.2
Ni crogen (N 2) Gas 0.0741 0.47S ( -J22) 0.2419 68-824 -J47.8 -Jl9 11.1 8S.6
-J44
Nitrous oxide (N20) Gas 0.117 .... . .. 0.2262 6140S .. -129 ..
Occane (C8H18) Liquid 0, F 4J.6 .... O.S2 60 . .... .... .
266.0 126.0
Oil (see castor oil, coal tar oil, cottonseed oil, fuel oil, fusel oil, light oil, linseed oil, machine oil, oil of citron, oil of juniper, oil of orange, oil of turpentine, olive oil, paraffin oil,
petroleum, and Chapter II.)
of cicron
of juniper
Liquid
Liquid
I SJ.2 I I 0.4J8
0.477
I 42 I
of orange Liquid 0.489
of turpentine Liquid 53. 7 . .. 0.411 32 14 J20 184
Olive oil Liquid S7.4 I ... . 0.471 44 68 S72
Osmium (Os) Solid E 1402 O.OJll 68-208 48921 9900
Oxalic acid (C2H204 2H20) Solid 0 IOJ.8 ( O.JJ8 0 372 JQ2C
0.416 100
Oxygen (02) Gas E 0.0847 o.J98 If -88.o-)I o.211s S5-40S -J60.4 -296.9 S.98 91.6
\ -79.2
Paraffin Solid S4S7 0.622 95-104 0.712 I 14014S l0013J I 662-806 6J70
Paraffin oil Liquid 0.S2 J2-212
Penrane n-CsH12 Liquid 0, F J8.8 97 IS4.4
Petroleum Liquid F 47SS O.Sll l69.81J6.4
Phenol (C5H50) (see also resin) Solid 66.8 lOS.6 J60 4S.O
Phosphorus (P) Solid E 113.8 0.1829 J2-124 111.6 SS0.4 9.0S
Piech (coal car) Solid F 6281 0.4S 60-212 I O.JS-0.4S 86-J02 J2S
Plaster Solid 90 0.20
Platinum (Pt) Solid E,M 13JS O.OJS9 68-2J72 Jl919 77SO 48.96
Porcelain Solid 14JIS6 0.26 S91742
Porcelain, refractory Solid R 0.2J 60-1200 21403000
I -JOI\
1 .....
a-A=alloy, E=elemenr, F=fuel or fuel compooent, !=insulation, M=metal, O=organic compound, R=refractory, S=solution.
C-sublimes.
,
,~c:o
Salt, rock
Samarium (Sm)
.
Solid.
Solid
I
I
E
13S
481
0.219 SH13 149S
2372-2552
Sand Solid 162 0.19S S9212
Sea w.are:- Liquid S 64 0.938 63.5
a-A=alloy, E=c:!cment, F=fudor fuel componear, !=insulation, M=rr..eral, O=organic compound, R=rcfractory, S=solution.
d-Manufacrnred under trade names: Bakelite, Redmanol, Condensice, etc.
e-Resins used in varnish making: Kauric, Congo, Zanzibar, and Manila copak
f-Sofrening point under load.
Selenium (Sc)
-
Sold
-E - -O 068
I 300
- --1
I -J06- 1 .
- - - - - -422- -------
6-428 o 1274
1
Sulfur (S)'
--- - - - - - - - - - - - - - - - --- --- - - - - - - - -pressure)
Solid E 119130 0.190 59130
fusion
--- ---- - -
1zation
0.2337 235-840 .
Sulfur dioxide (S02) 239 832.5 16.87h 651.6
Gas .. 0.1733 . . 0.36 . 122 0.1544 61396 -104 8 14
Sulfuric acid (H2S04) Liquid 115.8 I 0.336 32-212 50.9 640.4b
Talc Solid I 0 2092 68-208
Tal!ow, beef Solid 56.8-ii0.5 ( 0.79 79108\ 80-100
0.54 151-216{
Tantalum (Ta) Solid E 1035 0.043 2552 ... .
Tartaric acid (C4H606) Solid 5252 """
104 0.287 97 338
Tellurium (Te)
Thallium (Tl~
Thorium (Tb
I Solid
Solid
Solid
E
E
E
389
740
699
0.0483
0.0326
0.0276
59-212
68-212
32-212
845.6
575.6
2534
3000
13.1
.....
Tile, hollow 3350 >5432
Solid
Tin (Sn) Solid E,M
75
455
0.15
0.0551 70-228 I000,.,I
0.0758
482
2012
\
I
449.4 .4 4118 I I
Titanium (Ti) Solid E 28) 0.1125 32212 3263
Toluene (C7H8) Liquid 0 53.6 0.40 32-212 -133.6 230.5 150.3
Toluol (C5Hs) Liquid 0 0.490 149 230.5
Tufa
Tungsten (W)
Solid
Solid E, M 1202
0.33
J 0.0116
32-212
32-212 ) .. ~.
6152 10526 79
I 154.8
\ 0.0337 1832
Turpentine Liquid 53.6 0.42 32-212 318.8 133.J
Type metal
Uranium (U)
Solid
Solid
A
E 1167
0.0388
0.028
32-212
32-208
I
<.3344 6330
Vanadium (V) Solid F. 375 0.1153 32-212
Varnish (see resins) 3128 5430
Vegetable fiberboard ('Celotex") Solid [ 14.4 \ 0.171
0.279
-116
!09
\
I
I
Vermiculite (see mica) [
Vulcanite
W,der (H20) (see also sea water) Solid
Liquid
I 62.37
0.JJ12
0.480
68-212
<32 I 1.00
I 60 I I 32 I 212 I 144 I 970.2
Wood (see also redwood bark)
I
Solid 1956 0.33-0.67
I I
a-A=alloy, E=elemenc, F=fuel or fuel component, I=insuJatioo, M=metal, O=organic compound, R=refractory, S=solmion.
b-decomposes.
h-rran,formation from rhombic to monoclinic absorb~ S.06 Btu/lb.
___,___ ,______ - - - ------ - - - ---- - - - - heat range F heat range F beat range F pressnre)
Zioc (Zn) Solid E,M 445 f 0.0931 68-212 ) ~~-9 1663 46.8 7 58
\ 0.1040 572
Zinc chloride (ZnCl2) Solid 181.5 689 JJSO
Zinc oxide (ZnO) Solid 350 0.125 32-212 >3240
0.174 i6i;Oi: 1330b
, Zinc sulfate (ZnS04)
Zircon
LiquiCi
Solid
234
293 0.132 4622
Zirconium (Zr) Solid E 405 0.0660 32-212 3092 9122
a-A= alloy, E=element, F=fuel or fuel component, I= insulation, M=metal, O=organic compound, R=refractory, S=solutioo.
b-decomposes. ..
APPENDIX III*
i
\ ~"'
Cl'
I
,." ~.
Table III.1 Thermal properties of metals
/,~
'}
K',, .I
~'
Mean
specific
Mean
" , I
Latent specific Melt-
Mean Mean
specific Latent specific Melt-
Metal Composition Density, heat, Metal Composition Density, heat heatof heatof ing
heat of heat of I ing \
lb/cu ft 60- fusion, liquid, point, lb/cu ft 60- fusion, liquid, point,
melting Btu/lb Btu/lbF OF ! melting Btu/lb Btu/lbF OF
point, point.--..
Btu/lb 0 F. _/ Btu/lbF'
"
Aluminum Al 166.7 German Silver 60 Cu, 25 Zn, 15 Ni 0.109
0.248 169.0 0.26 1215 - 86.2 0.123 1850
Babbit, Lead Base 75 Pb, 15 Sb, 10 Sn Gold Au 1205 0.033
0.039 26.2 0.038 462 28.5 0.034 1945
Babbit, Tin Base 83.3 Sn, 8.4 Sb, 8.3 Cu 462 0.071 34.1 0.063 464
Iron 60-2786F Fe ~j)> 0.165. 89.0 0.150 2786
Bismuth Bi Lead Pb 708 0.032
612 0.033 18.5 0.035 518 10.0 0.034 621
Brass, Muntz Metal 60 Cu, 40 Zn 524 Linotype 86 Pb, 11 Sb, 3 Sn 0.036 21.5 0.036
0.105 69.0 0.125 1630 486
Magnesium Mg 108.6 0.272 83.7 0.266 1204
Manganese Mn 464 0.171 66.0 0.192 2246
Brass, Red 90 Cu, 10 Zn 546 0.104 86.5 0.115 1952 Mone! Metal 67 Ni, 28 Cu; Fe, Mn, Si 550 0.129 117.4
Brass, Yellow 67 Cu, 33 Zn 0.139 2415
528 0.105 71.0 0.123 1688
Bronze, Aluminum 90 Cu, 10 AI 510 0.126 98.5 0.125 1922 Nickel 60-2644F Ni 556 0.134 131.5
Bronze, Bearing 80 Cu, 10 Pb, 10 Sn 0.133 2644
556 0.095 79.9 0.109 1832
1
Silver Ag 665
Bronze, Bell Metal 78 Cu, 22 Sn 0.063 46.8 0.070 1762
540 0.100 76.3 0.119 1634 Solder, Bismuth 40 Pb, 20 Sn, 40 Bi 0.040 16.4 0.039 232
Solder, Plumbers' 50 Pb, 50 Sn 580 0.051 23.0 0.049 414
Bronze, Gun Metal 90 Cu, 10 Sn 550 Tin Sn 455 0.069 25.0 0.0637 450
0.107 84.2 0.106 1850
Bronze, Tobin 60 Cu, 39.2 Zn, 0.8 Sn Zinc Zn 445 0.107 48.0
525 0.107 73.5 0.124 1625 0.146 786
Copper Cu 559 0.104 91.0 0.111 '1982
Die Casting Metal 92 Al,8 Cu 176 0.236 163.0 0.241 1150
Die Casting Metal 80 Pb, 10 Sn, 10 Sb 0.038 17.5 0.037 600
Die Casting Metal 90 Sn, 4.5 Cu, 5.5 Sb 0.070 30.2 0.062 450
Die Casting Metal 87.3 Zn, 8.1 Sn, 4.1 Cu, 0.5 A 0.103 48.0 0.138 780
* Adapted from North American Combustion Handbook, published by North American Manufac-
turing Co., Cleveland, Ohio, 1952.
586
APPENDIX IV*
-----, I
----- --~--- ----- ----
CONVERSION FACTORS l
\
I
-ii
..
Table IV.1 Linear measure equivalents Table IV.1 Linear measure equivalents (continued)
Multiply by value
in table to obtain
these units -. angstrom centimeter .foot, inch kilometer meter micron mile millimeter yard
(A) (cm)' (ft) (in) (km) (m) () (mi) [mm) (yd)
Given in
these units !
angstrom 10-10 6.2X 10- 14 10-1 I. 0936 X 10- 1 0
3.937 x 10-9 10-
.
I 10-s 3.2808 x 10-10 10-13
(A)
centimeter
108 I 3. 2808 x 10-2 3.937 x 10- 1 10-s 10-2 104 6.2 x 10- 6 10 I . 0936 X 10-2
(cm)
foot
(ft}
3.48 x 109 @ I 12 3.048 x 10- 3.048 x 10- 1 3.048 X 105 I. 8939 x 10 - 3.048 x 102 3.333X 10- 1
micron
104 10- 3.2808 x 10-6 3.937 X 10-5 10-9 10-6 I 6.2137 x 10- 10 10-3 I .0936 x 10-6
()
millimeter 101 10-1 3.2808 x 10-3 3.937 x 10-2 10-6 10-3 103 6.2 x 10-7 I I .0936 x 10- 3
(mm)
yard ' 9.144 x 10-1 9.144 X 105 5.682X 10- 4 9.144 x 10+ 2 I
1.044 x 101q 9.144 X 10 1 3 36 9.144 x 10-4
(yd)
/
588
j
~
."
""
Given in
centimeter*
(cc or cm 3)
(ft 3) (in 3) (m3) (gal) (1)
ounces
(U.S. fluid oz) I
these units !
cubic centimeter 1 3.531 X 10- 5 6.103 x 10- 2 10-6 2.642 x 10-4 10-3 3.381x10- 2
(cc or cm 3 ~
cubic inch 5.787 x 10-4 4.329 x 10- 3 1.639 x 10- 2 5.541 x 10- 1
16.39 1 1.639 X 10- 5
(in 3)
cubic meter
(m3)
106 35.31 6.103 x 104 1 2.642 x 10 2 103 3.381 x 104
gallons (U.S.)
3. 785 X 103 1.337 x 10- 1 2.31 x 102 3.785 X 10-3 1 3.785 1.28 x 10 2
(gal)
liters
(I)
10a 3.531 x 10- 2 6.103 x 10- 1 10-a 2.642 x 10- 1 1 33.81
ounces
29.57 1.044 X 10- 3 1.805 2. 957 X 10- 5 7.812 x 10- 3 2.957X 10- 2 1
(U.S. fluid oz)
*Note: 1ml=1.000027 cc
,~--~---~_.....,.
..-
- -
Multiply by value
in table to obtain
these units --+ gram kilogram pound ton, ton, ounces
grains
(gm) (kg) (lb) long short (oz)
Given in
these units !
grain 1 6.48 x 10- 2 6.48 X 10- 5 1.429 x 10- 4 6.378 X 10-s 7.143X 10-s 2.286 X 10- 3
-
gram. (gm) 15.43 1 10-a 2.20 X 10- 3 9.84X 10- 7 _ 1.602 x 10- 5 3.527 x 10- 2
kilogram (kg) 1.543 x 104 10a 1 2.205 9.842 x 10-4 1. 102 x 10- 3 35.27
pound (lb) 7000 4.536 x 10 2 4.536 x 10- 1 1 4.464 x 10- 4 5.0 x 10-4 16
ton, long 1.568 x 10 7 1. 016 x 10 6 1. 016 x 103 2.24 x 10 3 1 1.12 3.584 x 10 4
ton, short 1.40 x 10 7 9.0718 X 10 5 9.072 x 10 2 2.00 x 10 3 8.929 x 10- 1 1 3.20 x 10 4
ounce (oz), 4.375 x 10 2 28.35
I
2.835 x 10- 2 6. 25 x 10- 2 2. 79 X 10- 5 3 .125 X 10- 5 1
.
o-"
Multiply by value
in table to obtain
these units -->
gm cm- 3 gm liter- 1 kg m- 3 lb. ft- 3 lb in- 3 lb US ga1- 1
Given in
these units l
gm cm- 3 1 10 3 lQ3 62.43 3.613 x 10- 2 8.345
plb \
I
I
P-
__J
I.
Multiply by value
in table to obtain
these units ----> dyne newton poundal pound force
(gm cm sec - 2) (kg m sec- 2 ) (lb ft sec- 2) (lb1)
Given in
these units l
dyne
1 10-5 7.233 x 10- 5 2.248 x 10- 6
(gm cm sec-2)
newton
(kg m sec- 2)
105 1 7.233 2.248 x 10- 1
poundal
1.3826 x 10 4 1. 3826 x 10- 1 1 3.108 x 10- 2
(lb. ft sec- 2 )
pound force
(lb1)
4.448 x 10 5 . 4.448 32.17 1
.,
"~
Multiply by value
in table to obtain
these units-+ I .
Btu cal ergs ft-lb hp-hr joule kcal kg-m kw-hr liter-atm I
Given in '
I I
I
these units l
l
Btu 2.52x10+ 2 J.055 x 10 10
I
I 7.7816 x 102 3.93 x 10- J.055 x 103 2.520 x 10- 1 1.0758 x 102 2.93 x 10- 10.41 I
I
cal 3.97 X 10-3 I 4. 184 x 107 3 086 J.558 x 10- 4. 184 10-3 4.267 x 10- 1 J.162XI0-6 4. 129 x 10-2 !
i
erg 9.478 x 10-7 2 .39 x 10- I
'
4 .376 x 10- 3. 125 x 10- 14 10-7 2 .39 x 10- 11 I .0197 X 10-8 2. 773 x 10- 14 9.869 x 10- 1 1
I
ft-lb J.285 X 10-3 3.241 x 10- 1 I 356 X 107 I 5.0505 x 10- 7 L356 3.241 x 10- 1.383 x 10- 1 3.766 x 10-1 1.338 x 10-2 I
I
I
hp-hr 2. 545 x 103 6.4162 x 10 5 2.6845 x 10 13 L98 X 106 I 2.6845 x 106 6.4162 x 10 2 2. 7375 x 10 5 7.455 x 10- 1 2.6494 x 10
I
joule 9.478 x 10- 2.39X 10- 1 10 7 7 .376 x 10- 1 3. 725 x 10- 7 I 2.39X10- J.0197 x 10- 1 2. 773 x 10-7 9.869 x 10- I
kcal 10 4.184 x 10 10 3.086 x 10 3 1.558 X 10-3 4.184 x 103 I 4.267 x 10 2 l.162X10- 3 41.29
3. 9657
I
kg-m 9.296.X 10-a 2.3438 9.8067 X 10 7 7.233 3.653X 10-6 9.8067 2.344 X 10-3 I 2. 724 x 10- 9.678 x 10- 2 [
kw-hr 3.4128 x 103 8. 6057 x 105 3.6 x 10 13 ,2 655; l~J 1341 3.6 x 10 6 8.6057 x 10 2 3.671 x 10 5 I 3.5534 x 104
I
9.604 x 10-2 j 1 0133 x 109 3.774 X 10- 5 1.0133 x 10 2 2.422 x 10- 2 2.815 x 10- I
liter-atm 24.218 74. 73 10.333
I
I therm = 100,000 Btu; also. I therm = 10 6 cal
Multiply by value
in table to obtain Column of Hg at O'C Column of H20 at I 5C
these units --+ atmosphere
kg cm- 2 kg m- 2 lb. ft- 2 lb. in- 2
(atm)
Given in
in mm ft in mm
these units l
atmosphere
I 1.0332 L033 X 104 2.1162 x 103 14.969 29.92 7.60 x 10 2 33.93 4.0714 x 10 2 1.034 x 10 4
(atm) '
kg cm- 2 9. 678 x 10- 1 I 10 4 2.048 x 103 14.22 28.96 7.335 x 10 2 32.84 3.9405 X 10 2 J.001 x 10 4
kg m- 2 9. 678 x 10- 10- I 2.048 x 10- 1 I 422 X 10-3 2.896 X 10-a 7.355 x 10-2 3.284 X 10-3 3 9405 x 10- 2 LOOI
lbft- 2 4. 725 x 10- 4.883 x 10- 4.883 I 6. 944 X 10-3 I 414 X 10- 2 3.591 x 10- 1 1.603 x 10-2 L924 X 10- 1 4.887
lb in- 2 6.804 x 10-2 7 .031 x 10-2 7 .031 x 10 2 144 I 2.036 5J. 71 2.309 27. 70 7 .037 X 10 2
Hg column,
f.316 X 10-3 1.3596 X 10-3 13.596 2. 7845 1.934 x 10-2 3.937 x 10- 2 I 4.464 x 10- 2 5.357 x 10- 1 13.61
mm
H20 column, 2.947 x 10- 2 3 .045 x 10- 2 3 .045 x 10 2 62 .372 4 .332 x 10- 1 8.819 x 10- 1 22.4 I 12 3 .048 X 10 2
ft
H20 column,
2.456 X 10-3 2.538 x 10-a 25.38 5.1977 3. 61 x 10- 2 7.349X 10-2 J.867 8.333 x 10-2 I 25.4
in
H20 column,
mm
9.67x10- 5 9. 991 x 10- 9.991 x 10- 1 2.046 x 10- 1 I l.421 X 10-3 2.893 X 10-3 7.349X 10-2 3.281 X 10-3 3.937x10- 2 I
Multiply by value
in table to obtain ,. ,,
these units---+ gm cm- 1 sec-I
cen'tipoise kg m- 1 ,sec- 1 lbm ft-I sect-I lbm ft-I hr-I 1b1 sec ft- 2
' (poise)
Given in
these units l
centipo!se . 1 10-3 10-2 6. 1191 x 10- 4 2. 4191 2. 0886 X 10- 5
.
j
I
Multiply by value I
in table to obtain
these units l l3tu cal ergs poundals lb1 watt
--
hr ft F sec cm K sec cm K sec ft F sec F mK
Given in
these units l
Btu
hr ft F
1 4:1365 x 10-3 1. 7307 x 10 5 6.9546 2. 1616 x 10- 1 1.7307
cal
2.4175 x 10 2 1 4.1840 x 10 7 1. 6813 x 10 3 52.256 4.1840 x 10 2
sec cm K
ergs
5. 7780 x fo- 6 2.3901 X 10- 8 1 4.0183 x 10- 5 1. 2489 x 10- 6 10- 5
sec cm K
poundals
sec ft F
1.4379X l0- 1 5.9479 x 10- 4 2.4886 x 10 4 1 3 .1081 x 10- 2 2. 4886 x 10- 1
1b1
4.6263 1. 9137 X 10- 2 8.0068 x 10 5 32.174 1 8.0068
sec F
watt
- - 5. 7780 x 10- 1 2.3901 x 10- 3 105 4.0183 1. 2489 x 10- 1 1
mK
-
~~~
LIST OF PRINCIPAL SYMBOLS
""0
-<
"'I
0
I -<
0
"'
rn '"I .X 'I
0 l:'- 0
-< 0 -<
00
ot:)
c<i A area
A constant, defined in Eq. (1.18)
'"-<
0
"'0I B particle mobility
-< Bi Biot number
x c molar concentration of entire solution
0
ot:) c Chvorinov's constant for solidification time
l:'-
00 c aperture conductance in vacuum
C'=1
C; molar concentration of species i in solution
CM maximum concentration
~
I 0
I
cm minimum concentration
0
-<
cp heat capacity at constant pressure
"'rn x '"0 ""I
0 cp molar heat capacity at constant pressure
l:'- -<
"'s 0
00
ot:)
-<
c, conductance of vapor trap in vacuum
0
C'l CV heat capacity at constant volume
CV molar heat capacity at constant volume
-4
c heat capacity per molecule
c speed oflight = 2.997902 x 10 10 cm/sec
-
i:i rn
-...,
a D,d diameter
-~ : "' -
i:: I ~
0 I D,D
rn
I 0 interdiffusion coefficient, or simply diffusion coefficient
0 i{j "'rn D* self-diffusion coefficient
~
s0
Db bubble diameter
De equivalent diameter, defined by Eq. (3.21)
D; intrinsic diffusion coefficient of species i
D;* self-diffusion coefficient of species i
Do frequency factor for diffusion coefficient
Dp particle diameter
J5P; average particle diameter of size fraction i
DT tracer determined diffusion coefficient
l5VS volume-surface mean diameter, defined. by Eq. (3.54)
'd crystal lattice dimension
d molecular collision diameter
E1 energy loss due to friction
599
/
------- - -- - ~- -
605
Chvorinov's rule, 332-333
Clausing factor, 561
Davenport, W. G., 165
Debye, P., 188
r through cylinders, 473-478
transient, 478, 482, 486
total, of a real body, 365
Emissivity
Collision cross section, 8 Decarburizing, 490-497 Diffusion into a falling liquid film, 521-~24 as function of view angle, 366
Collision integral Degassing, 74 Diffusion pump as function of wavelength, 366
definition, 11, 464 of molten aluminum, 74 Ho coefficient, 176 of gases, 390
graph, 464 of molten metals, 527-528 limiting forepressure, 175 of metal surfaces, 366-370
table, 12 of solid metals, 486 multi-stage, 174 of oxides, 366-368, 370
Combusion temperature, minimum, 570 Deissler, R. G., 201 speed of, 175-176 of refractories, 368, 370
Condensation coefficient, 564 De Laval nozzle, 160-162 ultimate pressure of, 176 reduced, 390
Conductance of vacuum system Derge, G., 459 vapor trap, 176, 177 Emittance
components, 170 Diffusion Diffosion through gases data, 367-369
definition, 170 in common liquids, 463 moving gas, 518-521 definition, 361
table, 171 stagnant gas, 515-518 End effects, 67
in gases, 463-473
Conduction heat transfer
'' c
'~
Fan laws, 155 Fluidized beds, 102 copision cross sections in, 466 for forced convection in tubes, 244-249
Fanning equation, 115 applications of, 106-107 I
diffusion coefficients in, 468 for natural convection, 252-256, 275
Fans elutriation in, 108 diffusion in, 463-467 in packed beds, 412-413
characteristic curve, 151 incipient fluidization in, 103 diffusion in binary gases, 464 on the casting side of interface, 346
' ~1 horsepower of, 151-153 particulate fluidization, 105 emissivity, 390 on the mold side of interface, 346
interaction with system, 153-157 Reynolds number in, 103 reduced emissivity, 390-393 over a flat plate, 212-215
pumping limit of, 154 Flux thermal conductivity of, 185-187 average, 215
static efficiency, 151-152 mass, 432 transmissivity, 396 local, 215
total efficiency, 151-152 molar, 432 viscosity of, 13 radiant, 387
Fermi energy, 191 units, 432 Gauvin, W. H., 558 total, 387
Fick's first law of diffusion, 432 Ford, H., 413 General energy equation total, across metal-mold interface, 349,
Fick's second law of diffusion, 535 Form drag, 87 components of, 229-233 346, 347
Film boiling, 258, 264-269 (see Friction factor for flow past a sphere) in cylindrical coordinates, 234, 235 units, 182
Film penetration theory, 544 Foundry sands in rectangular coordinates, 234, 235 volumetric, 406, 412-413
Film theory, 543 effect of, on solidification rates, 329 in spherical coordinates, 234, 235 Heat-transfer coefficients for
Flame front, rate of propagation, 415, 418, permeability of, 99 Glass, D.R., 167 flow over a flat plate, 251
419,425 thermal conductivity of, 203-205 Glass, thermal conductivity of, 195 flow past a cylinder, 249-250
Flame hardening, 328 Fourier number, 296-297 Graphite flow past a sphere, 250
Flemings, M. C., 349, 498, 499 Fourier theorem, 287 kinetics of dissolution, 541-542 Heat-transfer coefficients for natural con-
Flinn, R. A., 167 Fourier's law thermal conductivity of, 196 vection past
Flow around a sphere, 68 definition, 181, 183 Grashof number horizontal cylinders, 254, 275
momentum-flux distribution in, 69 in three dimensions, 184 for heat transfer, 223 horizontal surfaces, 256, 275
normal forces in, 69 Frenkel, J., 15 for mass transfer, 538 spheres, 254
pressure distribution in, 69 Friction factor Gray bodies, 369 vertical plates, 254, 275
velocity components, 69 dimensional analysis for, 77-79 Green, H. S., 18 Heat transfer with forced convection,
Flow between parallel plates, 41 for flow in rectangular ducts, 82 Griffiths, D. K., 262 in tubes, 207-212
average velocity in, 43 for flow in tubes, 79 Grossmann, M.A., 259 Heat transfer with natural convection, 220-
maximum velocity in, 43 for flow past flat plates, 83, 84 Grube solution, 480 226
velocity distribution in, 43 for flow past spheres, 88 Gurney-Lurie charts, 297 Heisler charts, 297
volume flow rate in, 43 for flow past submerged objects, 87 Higbie, R., 544
Flow coefficient, for met~rs, 128 for flow through packed beds, 94, 95 Hagen-Poiseuille law, 45 Hill, D.R., 262
Flow measurement, 123-134 Friction loss Head, 145 Hills, A. W. D., 349, 355, 356
(see Velocity meters, Head meters, and during enlargement and contraction, 117- actual, 147 Hirschfelder, J. o::J86, 465-
Area meters) 118 net positive suction, 146 Ho coefficient, 176
Flow nozzle, 127 in curved pipes, 121-123 pressure, 127 Hohlraum, 364
Flow over a flat plate, 59 in straight conduit, 114- H 4 theoretical, 147 Hole theory of liquids, 15
boundary layer in, 59-64 through valves and fittings, 118-120 Head meters, 126-131 Homogeneous boundary conditions, 288
drag force in, 64 Friction loss factor, 117 Heat diffusivity, 330 Homogenization, 497-502
Reynolds number for, 61, 64 Frictional heating, 114 Heat flux Hottel, H. C., 367, 368, 390
1'
'['
hydraulic radius Kinetic energy terms, 112-113 r average, 526 Nernst-Einstein equation, 443
in ductwork, 82 Kinetic theory of gases, 7, 185 definition, 525 extended, 451
in packed beds, 93 Kirkaldy, J. S., 513 local, 525 Newtonian fluid, 5
Hydrogen, removal from metal, 550 Kirkendall effect, 439 models, 542-545 Newtonian heating or cooling conditions for
Kirkwood, J. G., 18 Mass transfer, interphase validity, 291, 303
Inclusion flotation, 74 Kitaev, B. I., 412 film theory, 543 example problem, 292-293
Induction melting furnaces Knudsen diffusion, 468-470 two-resistance theory, 547-549 Newton's law of viscosity, 6
cooling water flow, 122 Knudsen effusion cells, 560-563 Mass transfer j-factor, 538-539 No-net-flux surfaces, 383
vacuum system design, 173 Kreith, F., 388, 392 Mass transfer resistances, 549 graph, 384
vaporization losses from, 517, 563, 565 Mata.no interface, 484 Noncircular conduits, 82
I
Insulation Ladles /j Mafano technique, 482 Non-Newtonian fluids, 31
critical radius of, 284 discharge coefficients for, 135 McAdams, W. H., 180, 256 Nozzle, converging
" effect of, 279, 283-285 time to discharge, 136 McKewan, W. M., 557, 558, 559 conditions for sonic flow, 159-160
Integral solution for solidification, 350-352 Laminar flow, 3 Mean free path, 8 mass flow rate from, 160
Interatomic spacing, 18 Langhaar, H., 67 Mechanical energy balance, 113 Nucleate boiling, 257, 269-270
table, 20 Laplace equation, 277 ~inimum ignition temperature, 570
Nusselt number
Interdiffusion, 441 Lead, diffusion in, 460 Mixed control of reactions average, 215
Interface resistance L1mnard-Jones potential, 10, 18, 464 gas-liquid, 547-551 definition, 211
between metal and mold, 344-349 Levitation melting gas-solid, 551-560 for laminar flow, 212
table, 349 heat transfer during, 403 Mixed mean temperature, 242 local, 215
Intrinsic diffusion coefficient, 432, 437 mass transfer during, 545 Mobility, 15
Intrinsic speed of vacuum pumps, 173 Lewis, W. K., 543 in an electrical field, 450-457, 504 Ogilvie, R. E., 513, 485
Iron ore sinter plant Li, Kun, 166 in an energy gradient, 442, 504 Oil, thermal conductivity of, 199
pressure drop across, 98, 99 Lightfoot, E. N., 186 in compounds, 443-444, 450-451 Oils, heat transfer in, 260-261
thermal profile in, 404 Limiting forepressure, 17 5 self-diffusion, 434 Orifice plate, 126, 129-130
Iron oxide Limons, R. A., 424 units of, 442, 451 Ostwald power law, 32
effect on slag viscosity, 28 Liquid metals Molar average velocity, 514 Overall mass transfer coefficients, 549
heating of, 415-417 diffusion in, 458-461 Molecular flow mechanics, 169-170 Oxidation
oxidation of, 418-419 structure of, 455-458 Molecular liquids, viscosity of, 1.6 parabolic, 503
Iron oxide reduction thermal conductivity in, 192, 197 Momentum
parabolic rate constants, 509
interface reaction control of, 557 Liquid phase control, example of, 565-566 balance, 36 Oxides
mixed control of, 558-560 Local mass transfer coefficients, 548 conservation of, 48
defect structures in, 449
Log mean concentrati.on, 517 convective, 49
diffusion coefficients in, 453, 455
j-factor Lorenz number, 191 flux,6, 37
thermal conductivity of, 188-190
heat transfer, 245 Love, T. J., 367 general equations, 47
data, 189
mass transfer, 538 Lucks, L. F., 205 viscous, 49
Jets, supersonic Momentum balance
Packed beds, heat-transfer coefficient in, 412
area of interaction with bath, 166 Mach number, 159 for flow between parallel plates, 42 Paschkis, V., 261
impact pressure, 165-167 Machlin, E. S., 563 for flow in a tube, 44
Peclet number, 526
length of core, 163 Manganese in a falling film, 37
Pehlke, R. D., 167
momentum of, 166-167 vaporization from molten steel, 566 Moore, M. R., 349, 355, 356
Pelletizing, 156
spreading of jet, 164 vaporization from solid alloys, 563 Morgan, E. R., 262
heat transfer during, 407, 414, 413
Johns, F. R., 163 Marangoni effect, 527 Penetration theory, 544
Jost, W., 478, 485 Marshall, W. R., 250 Nanda, C.R., 540 Permeability, 92
Mass transfer Natural convection data, 476, 477
Kapner, J. D., 166 to gas bubbles, 540 heat transfer, 254, 256, 275 d~finitions, 475, 476
Kattamis, T. F., 499. to spheres, 539 mass transfer, 538 specific permeability, 92
Kiriematic viscosity, 6 Mass transfer coefficient Navier-Stokes' equation, 47, 52, 59 Peterson, N. L., 485, 513
---------- ------------------------
Phonons, 188
centrifugal, 146 transition from laminar to turbulent in gas-liquid transfer, 525
mean free path of, 190
diffusion, 175 flow, 75, 80
Photon, 195 in laminar flow, 526, 537
efficiency of, 144-145 Rheopectic fluids, 33 in round tubes, 538
Pigford, R. L., 250, 524
electromagnetic, 150 Richardson, F. D., 455 over flat plates, 529-532
Pilling and Bedworth constant, 504
positive displacement, 143 Rohsenow, W. M., 241, 257, 270 Sherwood, T. K., 250
Pilling, N., 504
reciprocating, 143-144 Rotameter, 131-132 Sievert's law, 475
Pitot tubes, 124-126 rotary gear, 143-144 Roughness
Planck's equation, 364 Silica equivalence of aluminum oxide, 26
effect on heat transfer, 246 Siliconizing, 513
Poirier, D. R., 498 Quenching table, 81
Poise Similarity technique, 77-79
aluminum, 260 Ruddle, R. W., 349 Singh, S. N., 499
definition, 6 example problems, 276-277
units, 6, 7 Sintering, iron ore, 404, 414, 418, 419, 424
heat transfer coefficients in, 259-263 Salt solutions, heat transfer in, 260-261 Slags
.Polyalkylene-glycol, 259 heat transfer in, 257-263 Salts, molten diffusion in, 462
,'' Pqrnsity, effect on thermal conductivity, of steel sheet, 138 diffusion in, 460 structure of, 23-25
199-205 of strip, 261 thermal conductivity of, 197 thermal conductivity of, 197
Porous materials, thermal conductivity of, oils, 260-261 ~nd viscosity of, 23-30
199 salt solutions, 260-261 solidification in, 329-335 Smith, G. C., 165
Powder metallurgy thermal conductivity of, 203-205 Solidification
particle sizing, 89 Radiant heat transfer Sarofim, A. F., 367, 368, 390 against a chill, 335-340
powder rolling, 110 between infinite parallel plates, 370-372 Saunders, 0. H., 413 against a sand mold, 329-331
pressure drop for flow through compacts, electrical analog of, 379-387 Schack, A., 189, 194 effect of shape on, 331-335
100 within furnaces, 383, 384 Schmidt number effect of superheat on, 335
thermal conductivity of compacts, 201 Radiant thermal conductivity, 195 definition, 525 interface heat transfer during, 344-349
Practical parabolic scaling constant, 504, 506 Radiation from gases, 389--395 table of values, 531 rate, 331, 338, 345
Prandtl number Radiation heat-transfer coefficient, 387 Schneider, P. J., 398, 399 Sparrow, E. M., 267, 367
definition, 214 Radiation shields, 373 Schotte, W., 201
I Specific permeability, 92
table, 215 Radiosity, total, 361, 362 I Schroeder, D. L., 562, 563 Sphere (see Flow around a sphere)
Preheating of scrap, 424, 426
Pressure drop
for flow across a packed bed, 91, 94, 97
Ranz, W. E., 250
Rate-determining step, 430
Rational rate constant, 506
t
]~\
Segregation index, 489, 502
Self-diffusion, 433-437
Semiconductors, thermal conductivity of,
buoyant forces on, 71
form drag on, 71
friction drag on, 71
for flow through fittings, 120 Reduced temperature, 19
I
193-195 solidification time for, 339
for flow through a pipe; 79 Reduced viscosity, 19 Semi-infinite body, solution of heat flow Sprays, heat transfer with, 262, 263
for flow through a piping system, 115 Reduced volume, 19 equation for, 341 Stagnation conditions, 158
Pressure pouring, 139, 140 Reflectivity, 362 Semi-infinite plate, steady-state, conduction Stanton number, 245
Pressure vessel heat treatment, 327 through, 285-~89
Product solutions, 286, 321
Refractories, thermal conductivity of, 189 I Steady-state conduction through
Relative interface temperature during freez- Separation constant, 286 composite cylinder, 282-285
example of, 323-324 ing, graph, 342 Separation of variables, 285-287 infinite cylinder, 281-282
figure illustrating, 323 Relative position, 297 Severity of quench, 258-259 infinite flat plate, 278
for diffusion in solids, 488 Relative roughness, 80 Seybolt, A. U., 513
series composite wall, 278-280
Pseudoplastic fluids, viscosity, 32 Relative temperature, 297 Shape factor, 97
Steam ejectors, 178-179
Pumpdown time, 177 Reynik, R. J., 457, 458, 459 table, 96 Steel
Pumping limit, of fans, 154 Reynolds number, 78, 446 Shape factor for solidification in sand molds decarburization of, in electric furnace,
Pumping speed for flow over a flat plate, 61, 64 332-333 '
246-248
effective, I 72 for flow past submerged objects, 88 Sheet, heating of with radiation, 398-400 diffusion coefficients in liquid, 461
of vacuum pumps, 169 for flow through fluidized beds, 103 Sherby, 0. D., 447, 436
diffusion coefficients in solid, 447-448
of vacuum systems, 172 for flow through packed beds, 94, 96 Sherwood number
thermal conductivity of, 194
units, 169 for flow through a pipe, 46 definition, 525
Stefan-Boltzmann constant, 184, 365
Pumps (see Vacuum pumps) local, 64 prediction
Stefan-Boltzmann equation, 365
------------------------~-.-- --------------------------------------
l
infinite cylinders, 299 Thickness of solidified layer pumps, 172-177 Viscous momentum, 49
infinite plates, 279-299 in chill molds, 338-339 units of, 167-168 Voice, E.W., 99
spheres, 300-301 in sand molds, 331 /, Vacuum melting, vaporization during, 563- Void fraction, 100
Temporal mean velocity, 3 Thixotropic fluids, 33 567 at incipient fluidization, 103
Terminal velocity, 106 Thomas, C. G., 424 Vacuum pumps effect of packing on, 102
Themelis, N. J., 558 Throughput, 169 characteristic curve, 17 4, 176 Volume-surface mean diameter, 96, 97
Thermal boundary layer, 21Z, 215-219 Tin-plating, 510 diffusion, 174 Volumetric heat-transfer coefficient,
Thermal conduction, in packed beds, 420- Toor, H. L., 544 intrinsic speed of, 173 conversion factor for, 413
422 Torr, 167 limiting forepressure, 175 definition, 406
prediction of, 412-413 heat transfer in, 257-259
Von Karman integral relation, 66 thermal conductivity of, 197
Weidmann-Franz law, 191
Wakelin, D. H., 165 Wen, C. Y., 560
Wagner, C., 491, 504, 552, 555 Wetted perimeter, 82, 93
Wagner, J.B., 452, 455 Whitman, W., 543
Wall effect, 98 Wilke, c:R., l3
Ward, R. G., 566
Water Zener, C., 446
diffusion in, 463 Zuber, N., 271